[ { "title": "DiffuserCam: Lensless Single-exposure 3D Imaging", "authors": "Antipa et al.", "venue": "Optica", "year": 2018, "topic": "Lensless Imaging", "modality": "Coded optics", "status": "核心基线", "priority": "High", "task": "单次曝光无透镜三维重建", "method": "散射编码 PSF + 计算反卷积/正则化重建", "hardware": "Diffuser + conventional sensor", "dataset": "实验标定数据", "problem": "用散射片替代传统透镜，在单次曝光下恢复三维场景。", "contribution": "展示了低成本无透镜硬件和计算重建联合设计的可行性。", "limitation": "系统标定、噪声鲁棒性和大场景泛化仍是主要门槛。", "insight": "无透镜成像的关键不只是重建网络，而是编码设计、标定和逆问题可解性的整体配合。", "whyFollow": "适合作为 lensless imaging 方向的硬件-算法联合设计入口。", "openQuestion": "如何减少标定依赖，并让系统在动态场景和复杂深度范围内稳定工作。", "directionNote": "关注光学编码、PSF 标定、可制造性和端到端重建。", "tags": [ "lensless", "3D", "hardware-software co-design" ], "code": { "available": true, "url": "https://github.com/Waller-Lab/DiffuserCam" }, "url": "https://opg.optica.org/optica/fulltext.cfm?uri=optica-5-1-1&id=380522", "updated": "2026-05-20" }, { "title": "NeRF: Representing Scenes as Neural Radiance Fields", "authors": "Mildenhall et al.", "venue": "ECCV", "year": 2020, "topic": "Neural Radiance Fields", "modality": "Multi-view RGB", "status": "核心基线", "priority": "High", "task": "多视角新视角合成", "method": "隐式神经辐射场 + 体渲染优化", "hardware": "多视角 RGB 相机", "dataset": "Synthetic NeRF / LLFF", "problem": "从多视角图像学习连续三维场景表示并合成新视角。", "contribution": "用隐式神经场统一表示几何和外观，推动了神经渲染与逆问题结合。", "limitation": "原始方法训练慢，对动态场景、稀疏视角和真实传感噪声敏感。", "insight": "NeRF 把视觉重建从离散网格推向连续场表示，是计算成像和逆渲染的重要共同语言。", "whyFollow": "后续稀疏视角、动态场景、物理成像和显微/医学三维重建都绕不开这条线。", "openQuestion": "如何把真实成像噪声、曝光、散射和传感器响应纳入神经场训练。", "directionNote": "关注连续神经辐射场、稀疏视角重建、动态场景、物理成像模型和可编辑三维表示。", "tags": [ "neural fields", "inverse rendering", "3D" ], "code": { "available": true, "url": "https://github.com/bmild/nerf" }, "url": "https://www.matthewtancik.com/nerf", "updated": "2026-05-20" }, { "title": "3D Gaussian Splatting for Real-Time Radiance Field Rendering", "authors": "Kerbl et al.", "venue": "ACM TOG / SIGGRAPH", "year": 2023, "topic": "3D Gaussian Splatting", "modality": "Multi-view RGB", "status": "重点跟踪", "priority": "High", "task": "实时新视角合成和可交互三维表示", "method": "显式 3D Gaussian 表示 + 可微 splatting", "hardware": "多视角 RGB 相机 / GPU 渲染", "dataset": "Mip-NeRF360 / Tanks and Temples / Deep Blending", "problem": "在保持高质量新视角合成的同时显著提升训练和渲染速度。", "contribution": "用可优化的三维高斯替代纯隐式体渲染，带来实时渲染能力。", "limitation": "对动态、透明、强反射和物理一致成像的建模仍不完整。", "insight": "3DGS 让神经场研究从离线重建向可交互系统移动，适合接入计算成像采集链路。", "whyFollow": "它很可能成为三维计算成像原型系统的实时显示和评估层。", "openQuestion": "如何把低光、散射、显微、事件流等非标准观测纳入 Gaussian 优化。", "directionNote": "关注显式 Gaussian 表示、实时渲染、可微 splatting、动态场景和物理成像观测。", "tags": [ "3DGS", "real-time rendering", "neural rendering" ], "code": { "available": true, "url": "https://github.com/graphdeco-inria/gaussian-splatting" }, "url": "https://repo-sam.inria.fr/fungraph/3d-gaussian-splatting/", "updated": "2026-05-20" }, { "title": "Regularization by Denoising: Clarifications and New Interpretations", "authors": "Romano, Elad, Milanfar", "venue": "SIAM Journal on Imaging Sciences", "year": 2017, "topic": "Plug-and-Play Priors", "modality": "General inverse problems", "status": "核心基线", "priority": "High", "task": "把图像先验接入通用逆问题", "method": "RED / denoiser-induced regularization", "hardware": "通用成像系统", "dataset": "通用图像退化实验", "problem": "把强去噪器作为先验嵌入迭代逆问题求解。", "contribution": "为基于去噪器的正则化提供了清晰解释，影响后续 PnP 和 RED 方法。", "limitation": "理论条件与现代深度去噪器之间仍有差距。", "insight": "PnP/RED 的价值在于把物理 forward model 和学习先验解耦，便于快速迁移到新成像任务。", "whyFollow": "理解 diffusion priors 前，最好先掌握这一代 denoiser prior 的思想。", "openQuestion": "深度去噪器不满足经典假设时，如何给出可信收敛和误差解释。", "directionNote": "关注收敛理论、扩散先验、非凸求解和物理一致性。", "tags": [ "PnP", "RED", "inverse problems" ], "code": { "available": false, "url": "" }, "url": "https://epubs.siam.org/doi/10.1137/16M1102884", "updated": "2026-05-20" }, { "title": "Diffusion Posterior Sampling for General Noisy Inverse Problems", "authors": "Chung et al.", "venue": "ICLR", "year": 2023, "topic": "Diffusion Priors", "modality": "General inverse problems", "status": "重点跟踪", "priority": "High", "task": "带噪逆问题后验采样", "method": "预训练 diffusion prior + measurement consistency guidance", "hardware": "通用成像系统", "dataset": "多类图像退化任务", "problem": "将预训练扩散模型用于退化、带噪观测下的后验采样。", "contribution": "提供了把强生成先验接入多类逆问题的通用路径。", "limitation": "采样成本高，物理约束和不确定性校准仍需谨慎验证。", "insight": "扩散先验把逆问题从单点估计推进到后验采样，但必须警惕 hallucination 和测量不一致。", "whyFollow": "这是当前计算成像与生成模型结合最活跃的方向之一。", "openQuestion": "如何在真实硬件噪声和未知退化下保证结果可信。", "directionNote": "关注可控采样、快速求解、可信不确定性和真实成像退化。", "tags": [ "diffusion", "posterior sampling", "Bayesian" ], "code": { "available": true, "url": "https://github.com/DPS2022/diffusion-posterior-sampling" }, "url": "https://openreview.net/forum?id=OnD9zGAGT0k", "updated": "2026-05-20" }, { "title": "Score-Based Generative Modeling through Stochastic Differential Equations", "authors": "Song et al.", "venue": "ICLR", "year": 2021, "topic": "Diffusion Priors", "modality": "General image priors", "status": "理论入口", "priority": "High", "task": "连续时间 score-based generative modeling", "method": "SDE formulation + score matching + predictor-corrector sampling", "hardware": "通用 GPU 训练", "dataset": "CIFAR-10 / CelebA-HQ / LSUN 等", "problem": "用随机微分方程统一 score-based 生成模型的训练和采样。", "contribution": "给后续 diffusion inverse problem 方法提供了核心生成先验框架。", "limitation": "本身不是成像逆问题论文，需要结合 measurement operator 才能服务计算成像。", "insight": "理解 SDE/score matching 能帮助判断 diffusion restoration 方法是否真的尊重观测模型。", "whyFollow": "它是读懂 diffusion posterior sampling、DDRM 和后续扩散先验方法的基础。", "openQuestion": "如何把连续时间采样过程压缩到可用于实时成像系统的成本。", "directionNote": "关注 score prior、posterior sampling、fast samplers 和 uncertainty quantification。", "tags": [ "score model", "SDE", "generative prior" ], "code": { "available": true, "url": "https://github.com/yang-song/score_sde" }, "url": "https://openreview.net/forum?id=PxTIG12RRHS", "updated": "2026-05-20" }, { "title": "Deep Image Prior", "authors": "Ulyanov, Vedaldi, Lempitsky", "venue": "CVPR", "year": 2018, "topic": "Implicit Priors", "modality": "General inverse problems", "status": "核心基线", "priority": "Medium", "task": "无训练数据图像恢复", "method": "untrained convolutional network as implicit image prior", "hardware": "通用成像系统", "dataset": "denoising / inpainting / super-resolution", "problem": "不依赖外部训练集，仅用网络结构本身作为图像先验。", "contribution": "证明未训练网络的结构偏置能在多种逆问题中产生有效正则化。", "limitation": "需要早停，速度慢，对复杂真实退化的能力有限。", "insight": "DIP 是理解“网络结构本身也是先验”的关键论文。", "whyFollow": "适合和 PnP、RED、diffusion prior 放在一起比较先验来源。", "openQuestion": "如何稳定控制过拟合，并把隐式先验扩展到视频、三维和物理传感器数据。", "directionNote": "关注无需训练数据的先验、早停机制和物理观测一致性。", "tags": [ "implicit prior", "untrained network", "inverse problems" ], "code": { "available": true, "url": "https://github.com/DmitryUlyanov/deep-image-prior" }, "url": "https://dmitryulyanov.github.io/deep_image_prior", "updated": "2026-05-20" }, { "title": "Fourier Ptychographic Microscopy", "authors": "Zheng, Horstmeyer, Yang", "venue": "Nature Photonics", "year": 2013, "topic": "Computational Microscopy", "modality": "Microscopy", "status": "核心基线", "priority": "High", "task": "大视场高分辨率相位恢复", "method": "多角度照明 + Fourier domain stitching + phase retrieval", "hardware": "LED array microscope", "dataset": "显微实验采集", "problem": "突破传统显微镜视场和分辨率之间的折中。", "contribution": "通过多角度照明和相位恢复重建高分辨率大视场图像。", "limitation": "采集速度、系统误差校正和厚样本建模仍影响应用。", "insight": "FPM 是计算成像里“改变采集方式换取可计算分辨率”的代表范式。", "whyFollow": "它连接了硬件照明设计、相位恢复、显微成像和高通量生物应用。", "openQuestion": "如何处理厚样本、多散射和实时采集。", "directionNote": "关注高通量显微、相位恢复、系统校正和生物医学场景。", "tags": [ "ptychography", "phase retrieval", "microscopy" ], "code": { "available": false, "url": "" }, "url": "https://www.nature.com/articles/nphoton.2013.187", "updated": "2026-05-20" }, { "title": "Event-based Vision: A Survey", "authors": "Gallego et al.", "venue": "IEEE TPAMI", "year": 2022, "topic": "Event-based Imaging", "modality": "Event camera", "status": "综述入口", "priority": "Medium", "task": "事件相机研究路线综述", "method": "taxonomy of event representations, reconstruction and perception", "hardware": "Dynamic Vision Sensor", "dataset": "事件视觉公开数据集", "problem": "梳理事件相机的传感机制、表示学习、重建和视觉任务。", "contribution": "为事件成像研究提供系统入口，覆盖低延迟和高动态范围场景。", "limitation": "综述本身不解决事件流噪声、标定和跨传感器泛化问题。", "insight": "事件相机不是普通视频的高帧率版本，而是异步稀疏视觉信号。", "whyFollow": "适合作为进入 neuromorphic computational imaging 的地图。", "openQuestion": "事件流如何和 frame、depth、IMU、spike sensor 等多模态观测稳定融合。", "directionNote": "关注异步传感、动态重建、低功耗视觉和神经形态硬件。", "tags": [ "event camera", "survey", "neuromorphic" ], "code": { "available": false, "url": "" }, "url": "https://ieeexplore.ieee.org/document/9138762", "updated": "2026-05-20" }, { "title": "Single-pixel Imaging via Compressive Sampling", "authors": "Duarte et al.", "venue": "IEEE Signal Processing Magazine", "year": 2008, "topic": "Compressive Imaging", "modality": "Single-pixel camera", "status": "核心基线", "priority": "Medium", "task": "单像素压缩成像", "method": "coded measurements + compressive sensing reconstruction", "hardware": "DMD + single-pixel detector", "dataset": "实验编码测量", "problem": "用单点探测器和编码测量恢复二维图像。", "contribution": "把压缩感知和成像硬件结合，形成单像素成像代表范式。", "limitation": "采样效率、运动场景和高分辨率扩展存在挑战。", "insight": "单像素成像的价值在于把传感器阵列成本转移到编码和重建算法上。", "whyFollow": "适合关注非可见波段、低成本探测器和压缩采样硬件的研究。", "openQuestion": "如何在动态场景中减少测量次数并保持鲁棒重建。", "directionNote": "关注编码设计、快速采样、低光照和非可见波段成像。", "tags": [ "compressive sensing", "coded measurement", "hardware" ], "code": { "available": false, "url": "" }, "url": "https://ieeexplore.ieee.org/document/4472247", "updated": "2026-05-20" }, { "title": "Learning to See in the Dark", "authors": "Chen et al.", "venue": "CVPR", "year": 2018, "topic": "Low-light Computational Photography", "modality": "Raw sensor", "status": "复现候选", "priority": "Medium", "task": "极低照度 raw 图像恢复", "method": "raw-to-RGB neural image processing pipeline", "hardware": "Sony/Fuji raw sensor", "dataset": "See-in-the-Dark dataset", "problem": "在极低照度下从短曝光 raw 数据恢复可用图像。", "contribution": "展示了 raw 域学习和真实传感器数据对低光增强的重要性。", "limitation": "跨相机泛化、运动模糊和真实噪声建模仍有限。", "insight": "低光增强不能只在 sRGB 上做，raw domain 才接近真实成像链路。", "whyFollow": "适合做 raw pipeline、噪声模型和端侧低光成像的基线。", "openQuestion": "如何处理运动、不同 ISP、不同传感器之间的泛化。", "directionNote": "关注 raw pipeline、噪声模型、真实数据集和端侧部署。", "tags": [ "raw", "low-light", "computational photography" ], "code": { "available": true, "url": "https://github.com/cchen156/Learning-to-See-in-the-Dark" }, "url": "https://cchen156.github.io/SID.html", "updated": "2026-05-20" }, { "title": "Neural Holography with Camera-in-the-loop Training", "authors": "Peng et al.", "venue": "ACM TOG / SIGGRAPH Asia", "year": 2020, "topic": "Computational Displays", "modality": "Holographic display", "status": "复现候选", "priority": "Medium", "task": "真实系统闭环全息显示优化", "method": "camera-in-the-loop calibration + learned hologram generation", "hardware": "SLM + camera feedback display bench", "dataset": "camera-captured display calibration", "problem": "缩小模拟波前传播模型和真实全息显示系统之间的差距。", "contribution": "用相机闭环训练把硬件误差纳入全息图生成。", "limitation": "依赖具体硬件标定，系统搭建和复现实验成本较高。", "insight": "真实计算成像系统往往败在 model mismatch，camera-in-the-loop 是直接处理 mismatch 的有效范式。", "whyFollow": "对硬件闭环优化、显示成像和 differentiable optics 都有参考价值。", "openQuestion": "如何减少每台设备的重新标定成本，并推广到动态和宽视场显示。", "directionNote": "关注 differentiable optics、hardware-in-the-loop 和真实系统误差建模。", "tags": [ "holography", "camera-in-the-loop", "differentiable optics" ], "code": { "available": true, "url": "https://github.com/computational-imaging/neural-holography" }, "url": "https://www.computationalimaging.org/publications/neuralholography/", "updated": "2026-05-20" }, { "title": "Densely Connected Convolutional Networks", "authors": "Gao Huang, Zhuang Liu, Kilian Q. Weinberger", "venue": "CVPR", "year": 2016, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "High", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "Recent work has shown that convolutional networks can be substantially deeper, more accurate, and efficient to train if they contain shorter connections between layers close to the input and those close to the output. In this paper, we embrace this observation and introduce the Dense Convolutional Network (DenseNet), which connects each layer to every other layer in a feed-forward fashion. Whereas traditional convolutional networks with L layers have L connections—one between each layer and its subsequent layer—our network has L(L+1)/2 direct connections. For each layer, the feature-maps of all preceding layers are used as inputs, and its own feature-maps are used as inputs into all subsequent layers. DenseNets have several compelling advantages: they alleviate the vanishing-gradient problem, strengthen feature propagation, encourage feature reuse, and substantially reduce the number of parameters. We evaluate our proposed architecture on four highly competitive object recognition benchmark tasks (CIFAR-10, CIFAR-100, SVHN, and ImageNet). DenseNets obtain significant improvements over the state-of-the-art on most of them, whilst requiring less memory and computation to achieve high performance. Code and pre-trained models are available at https://github.com/liuzhuang13/DenseNet.", "motivation": "DenseNets have several compelling advantages: they alleviate the vanishing-gradient problem, strengthen feature propagation, encourage feature reuse, and substantially reduce the number of parameters.", "implementation": "In this paper, we embrace this observation and introduce the Dense Convolutional Network (DenseNet), which connects each layer to every other layer in a feed-forward fashion.", "application": "We evaluate our proposed architecture on four highly competitive object recognition benchmark tasks (CIFAR-10, CIFAR-100, SVHN, and ImageNet).", "problem": "DenseNets have several compelling advantages: they alleviate the vanishing-gradient problem, strengthen feature propagation, encourage feature reuse, and substantially reduce the number of parameters.", "contribution": "In this paper, we embrace this observation and introduce the Dense Convolutional Network (DenseNet), which connects each layer to every other layer in a feed-forward fashion.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Densely Connected Convolutional Networks 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 43,024；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Datasets / Benchmarks", "award-winning", "Best Paper" ], "code": { "available": true, "url": "https://github.com/liuzhuang13/DenseNet" }, "url": "https://arxiv.org/pdf/1608.06993", "updated": "2026-05-20", "topCitedRank": 1, "citationCount": 43024, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/5694e46284460a648fe29117cbc55f6c9be3fa3c", "doi": "10.1109/CVPR.2017.243", "arxiv": "1608.06993", "awards": [ { "conference": "CVPR", "year": 2017, "award": "Best Paper", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ] }, { "title": "Swin Transformer: Hierarchical Vision Transformer using Shifted Windows", "authors": "Ze Liu, Yutong Lin, Yue Cao et al.", "venue": "ICCV", "year": 2021, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "High", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "This paper presents a new vision Transformer, called Swin Transformer, that capably serves as a general-purpose backbone for computer vision. Challenges in adapting Transformer from language to vision arise from differences between the two domains, such as large variations in the scale of visual entities and the high resolution of pixels in images compared to words in text. To address these differences, we propose a hierarchical Transformer whose representation is computed with Shifted windows. The shifted windowing scheme brings greater efficiency by limiting self-attention computation to non-overlapping local windows while also allowing for cross-window connection. This hierarchical architecture has the flexibility to model at various scales and has linear computational complexity with respect to image size. These qualities of Swin Transformer make it compatible with a broad range of vision tasks, including image classification (87.3 top-1 accuracy on ImageNet-1K) and dense prediction tasks such as object detection (58.7 box AP and 51.1 mask AP on COCO test-dev) and semantic segmentation (53.5 mIoU on ADE20K val). Its performance surpasses the previous state-of-the-art by a large margin of +2.7 box AP and +2.6 mask AP on COCO, and +3.2 mIoU on ADE20K, demonstrating the potential of Transformer-based models as vision backbones. The hierarchical design and the shifted window approach also prove beneficial for all-MLP architectures. The code and models are publicly available at https://github.com/microsoft/Swin-Transformer.", "motivation": "Challenges in adapting Transformer from language to vision arise from differences between the two domains, such as large variations in the scale of visual entities and the high resolution of pixels in images compared to words in text.", "implementation": "To address these differences, we propose a hierarchical Transformer whose representation is computed with Shifted windows.", "application": "These qualities of Swin Transformer make it compatible with a broad range of vision tasks, including image classification (87.3 top-1 accuracy on ImageNet-1K) and dense prediction tasks such as object detection (58.7 box AP and 51.1 mask AP on COCO test-dev) and semantic segmentation (53.5 mIoU on ADE20K val).", "problem": "Challenges in adapting Transformer from language to vision arise from differences between the two domains, such as large variations in the scale of visual entities and the high resolution of pixels in images compared to words in text.", "contribution": "To address these differences, we propose a hierarchical Transformer whose representation is computed with Shifted windows.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Swin Transformer: Hierarchical Vision Transformer using Shifted Windows 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 32,038；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "ICCV", "Datasets / Benchmarks", "award-winning", "Best Paper" ], "code": { "available": true, "url": "https://github.com/microsoft/Swin-Transformer" }, "url": "http://arxiv.org/pdf/2103.14030", "updated": "2026-05-20", "topCitedRank": 2, "citationCount": 32038, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/c8b25fab5608c3e033d34b4483ec47e68ba109b7", "doi": "10.1109/ICCV48922.2021.00986", "arxiv": "2103.14030", "awards": [ { "conference": "ICCV", "year": 2021, "award": "Best Paper", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ] }, { "title": "Focal Loss for Dense Object Detection", "authors": "Tsung-Yi Lin, Priya Goyal, Ross B. Girshick et al.", "venue": "ICCV", "year": 2017, "topic": "Object Detection", "modality": "RGB / multi-scale visual features", "status": "CV 顶会高引 Top 100", "priority": "High", "task": "目标检测、开放集检测或检测损失设计", "method": "检测框架、特征金字塔、transformer detector 或检测损失", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "The highest accuracy object detectors to date are based on a two-stage approach popularized by R-CNN, where a classifier is applied to a sparse set of candidate object locations. In contrast, one-stage detectors that are applied over a regular, dense sampling of possible object locations have the potential to be faster and simpler, but have trailed the accuracy of two-stage detectors thus far. In this paper, we investigate why this is the case. We discover that the extreme foreground-background class imbalance encountered during training of dense detectors is the central cause. We propose to address this class imbalance by reshaping the standard cross entropy loss such that it down-weights the loss assigned to well-classified examples. Our novel Focal Loss focuses training on a sparse set of hard examples and prevents the vast number of easy negatives from overwhelming the detector during training. To evaluate the effectiveness of our loss, we design and train a simple dense detector we call RetinaNet. Our results show that when trained with the focal loss, RetinaNet is able to match the speed of previous one-stage detectors while surpassing the accuracy of all existing state-of-the-art two-stage detectors. Code is at: https://github.com/facebookresearch/Detectron.", "motivation": "The highest accuracy object detectors to date are based on a two-stage approach popularized by R-CNN, where a classifier is applied to a sparse set of candidate object locations.", "implementation": "We propose to address this class imbalance by reshaping the standard cross entropy loss such that it down-weights the loss assigned to well-classified examples.", "application": "To evaluate the effectiveness of our loss, we design and train a simple dense detector we call RetinaNet.", "problem": "The highest accuracy object detectors to date are based on a two-stage approach popularized by R-CNN, where a classifier is applied to a sparse set of candidate object locations.", "contribution": "We propose to address this class imbalance by reshaping the standard cross entropy loss such that it down-weights the loss assigned to well-classified examples.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Focal Loss for Dense Object Detection 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Object Detection 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 31,465；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何在开放世界、长尾类别和实时部署中保持稳定检测性能。", "directionNote": "该方向的高引用工作集中体现了 Object Detection 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "ICCV", "Object Detection", "award-winning", "Best Student" ], "code": { "available": true, "url": "https://github.com/facebookresearch/Detectron" }, "url": "http://arxiv.org/pdf/1708.02002", "updated": "2026-05-20", "topCitedRank": 3, "citationCount": 31465, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/79cfb51a51fc093f66aac8e858afe2e14d4a1f20", "doi": "10.1109/ICCV.2017.324", "arxiv": "1708.02002", "awards": [ { "conference": "ICCV", "year": 2017, "award": "Best Student Paper", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ] }, { "title": "Feature Pyramid Networks for Object Detection", "authors": "Tsung-Yi Lin, Piotr Dollár, Ross B. Girshick et al.", "venue": "CVPR", "year": 2016, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "High", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "Feature pyramids are a basic component in recognition systems for detecting objects at different scales. But pyramid representations have been avoided in recent object detectors that are based on deep convolutional networks, partially because they are slow to compute and memory intensive. In this paper, we exploit the inherent multi-scale, pyramidal hierarchy of deep convolutional networks to construct feature pyramids with marginal extra cost. A top-down architecture with lateral connections is developed for building high-level semantic feature maps at all scales. This architecture, called a Feature Pyramid Network (FPN), shows significant improvement as a generic feature extractor in several applications. Using a basic Faster R-CNN system, our method achieves state-of-the-art single-model results on the COCO detection benchmark without bells and whistles, surpassing all existing single-model entries including those from the COCO 2016 challenge winners. In addition, our method can run at 5 FPS on a GPU and thus is a practical and accurate solution to multi-scale object detection. Code will be made publicly available.", "motivation": "Using a basic Faster R-CNN system, our method achieves state-of-the-art single-model results on the COCO detection benchmark without bells and whistles, surpassing all existing single-model entries including those from the COCO 2016 challenge winners.", "implementation": "A top-down architecture with lateral connections is developed for building high-level semantic feature maps at all scales.", "application": "Feature pyramids are a basic component in recognition systems for detecting objects at different scales.", "problem": "Using a basic Faster R-CNN system, our method achieves state-of-the-art single-model results on the COCO detection benchmark without bells and whistles, surpassing all existing single-model entries including those from the COCO 2016 challenge winners.", "contribution": "A top-down architecture with lateral connections is developed for building high-level semantic feature maps at all scales.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Feature Pyramid Networks for Object Detection 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 26,678；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Datasets / Benchmarks" ], "code": { "available": false, "url": "" }, "url": "https://arxiv.org/pdf/1612.03144", "updated": "2026-05-20", "topCitedRank": 4, "citationCount": 26678, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/2a94c84383ee3de5e6211d43d16e7de387f68878", "doi": "10.1109/CVPR.2017.106", "arxiv": "1612.03144" }, { "title": "High-Resolution Image Synthesis with Latent Diffusion Models", "authors": "Robin Rombach, A. Blattmann, Dominik Lorenz et al.", "venue": "CVPR", "year": 2021, "topic": "Vision-Language / Foundation Models", "modality": "Image-text / foundation model", "status": "CV 顶会高引 Top 100", "priority": "High", "task": "视觉-语言理解、开放词表识别或基础模型适配", "method": "大规模预训练、多模态对齐或提示式视觉建模", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "By decomposing the image formation process into a sequential application of denoising autoencoders, diffusion models (DMs) achieve state-of-the-art synthesis results on image data and beyond. Additionally, their formulation allows for a guiding mechanism to control the image generation process without retraining. However, since these models typically operate directly in pixel space, optimization of powerful DMs often consumes hundreds of GPU days and inference is expensive due to sequential evaluations. To enable DM training on limited computational resources while retaining their quality and flexibility, we apply them in the latent space of powerful pretrained autoencoders. In contrast to previous work, training diffusion models on such a representation allows for the first time to reach a near-optimal point between complexity reduction and detail preservation, greatly boosting visual fidelity. By introducing cross-attention layers into the model architecture, we turn diffusion models into powerful and flexible generators for general conditioning inputs such as text or bounding boxes and high-resolution synthesis becomes possible in a convolutional manner. Our latent diffusion models (LDMs) achieve new state of the art scores for image inpainting and class-conditional image synthesis and highly competitive performance on various tasks, including unconditional image generation, text-to-image synthesis, and super-resolution, while significantly reducing computational requirements compared to pixel-based DMs.", "motivation": "However, since these models typically operate directly in pixel space, optimization of powerful DMs often consumes hundreds of GPU days and inference is expensive due to sequential evaluations.", "implementation": "By introducing cross-attention layers into the model architecture, we turn diffusion models into powerful and flexible generators for general conditioning inputs such as text or bounding boxes and high-resolution synthesis becomes possible in a convolutional manner.", "application": "By decomposing the image formation process into a sequential application of denoising autoencoders, diffusion models (DMs) achieve state-of-the-art synthesis results on image data and beyond.", "problem": "However, since these models typically operate directly in pixel space, optimization of powerful DMs often consumes hundreds of GPU days and inference is expensive due to sequential evaluations.", "contribution": "By introducing cross-attention layers into the model architecture, we turn diffusion models into powerful and flexible generators for general conditioning inputs such as text or bounding boxes and high-resolution synthesis becomes possible in a convolutional manner.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "High-Resolution Image Synthesis with Latent Diffusion Models 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Vision-Language / Foundation Models 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 24,464；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何降低标注和算力成本，同时保持可控性、可解释性和领域泛化。", "directionNote": "该方向的高引用工作集中体现了 Vision-Language / Foundation Models 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Vision-Language / Foundation Models" ], "code": { "available": false, "url": "" }, "url": "https://arxiv.org/pdf/2112.10752", "updated": "2026-05-20", "topCitedRank": 5, "citationCount": 24464, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/c10075b3746a9f3dd5811970e93c8ca3ad39b39d", "doi": "10.1109/CVPR52688.2022.01042", "arxiv": "2112.10752" }, { "title": "CBAM: Convolutional Block Attention Module", "authors": "Sanghyun Woo, Jongchan Park, Joon-Young Lee et al.", "venue": "ECCV", "year": 2018, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "High", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "We propose Convolutional Block Attention Module (CBAM), a simple yet effective attention module for feed-forward convolutional neural networks. Given an intermediate feature map, our module sequentially infers attention maps along two separate dimensions, channel and spatial, then the attention maps are multiplied to the input feature map for adaptive feature refinement. Because CBAM is a lightweight and general module, it can be integrated into any CNN architectures seamlessly with negligible overheads and is end-to-end trainable along with base CNNs. We validate our CBAM through extensive experiments on ImageNet-1K, MS COCO detection, and VOC 2007 detection datasets. Our experiments show consistent improvements in classification and detection performances with various models, demonstrating the wide applicability of CBAM. The code and models will be publicly available.", "motivation": "We propose Convolutional Block Attention Module (CBAM), a simple yet effective attention module for feed-forward convolutional neural networks.", "implementation": "Given an intermediate feature map, our module sequentially infers attention maps along two separate dimensions, channel and spatial, then the attention maps are multiplied to the input feature map for adaptive feature refinement.", "application": "We validate our CBAM through extensive experiments on ImageNet-1K, MS COCO detection, and VOC 2007 detection datasets.", "problem": "We propose Convolutional Block Attention Module (CBAM), a simple yet effective attention module for feed-forward convolutional neural networks.", "contribution": "Given an intermediate feature map, our module sequentially infers attention maps along two separate dimensions, channel and spatial, then the attention maps are multiplied to the input feature map for adaptive feature refinement.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "CBAM: Convolutional Block Attention Module 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 23,304；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "ECCV", "Datasets / Benchmarks" ], "code": { "available": false, "url": "" }, "url": "https://www.semanticscholar.org/paper/de95601d9e3b20ec51aa33e1f27b1880d2c44ef2", "updated": "2026-05-20", "topCitedRank": 6, "citationCount": 23304, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/de95601d9e3b20ec51aa33e1f27b1880d2c44ef2", "doi": "10.1007/978-3-030-01234-2_1", "arxiv": "1807.06521" }, { "title": "Image-to-Image Translation with Conditional Adversarial Networks", "authors": "Phillip Isola, Jun-Yan Zhu, Tinghui Zhou et al.", "venue": "CVPR", "year": 2016, "topic": "Generative Vision", "modality": "Image generation / editing", "status": "CV 顶会高引 Top 100", "priority": "High", "task": "图像生成、编辑、翻译或高保真合成", "method": "GAN、diffusion、transformer 或条件生成模型", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "We investigate conditional adversarial networks as a general-purpose solution to image-to-image translation problems. These networks not only learn the mapping from input image to output image, but also learn a loss function to train this mapping. This makes it possible to apply the same generic approach to problems that traditionally would require very different loss formulations. We demonstrate that this approach is effective at synthesizing photos from label maps, reconstructing objects from edge maps, and colorizing images, among other tasks. Moreover, since the release of the pix2pix software associated with this paper, hundreds of twitter users have posted their own artistic experiments using our system. As a community, we no longer hand-engineer our mapping functions, and this work suggests we can achieve reasonable results without handengineering our loss functions either.", "motivation": "This makes it possible to apply the same generic approach to problems that traditionally would require very different loss formulations.", "implementation": "These networks not only learn the mapping from input image to output image, but also learn a loss function to train this mapping.", "application": "We demonstrate that this approach is effective at synthesizing photos from label maps, reconstructing objects from edge maps, and colorizing images, among other tasks.", "problem": "This makes it possible to apply the same generic approach to problems that traditionally would require very different loss formulations.", "contribution": "These networks not only learn the mapping from input image to output image, but also learn a loss function to train this mapping.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Image-to-Image Translation with Conditional Adversarial Networks 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Generative Vision 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 22,267；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何平衡生成质量、可控性、版权风险和真实任务可靠性。", "directionNote": "该方向的高引用工作集中体现了 Generative Vision 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Generative Vision" ], "code": { "available": false, "url": "" }, "url": "https://arxiv.org/pdf/1611.07004", "updated": "2026-05-20", "topCitedRank": 7, "citationCount": 22267, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/8acbe90d5b852dadea7810345451a99608ee54c7", "doi": "10.1109/CVPR.2017.632", "arxiv": "1611.07004" }, { "title": "End-to-End Object Detection with Transformers", "authors": "Nicolas Carion, Francisco Massa, Gabriel Synnaeve et al.", "venue": "ECCV", "year": 2020, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "High", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "We present a new method that views object detection as a direct set prediction problem. Our approach streamlines the detection pipeline, effectively removing the need for many hand-designed components like a non-maximum suppression procedure or anchor generation that explicitly encode our prior knowledge about the task. The main ingredients of the new framework, called DEtection TRansformer or DETR, are a set-based global loss that forces unique predictions via bipartite matching, and a transformer encoder-decoder architecture. Given a fixed small set of learned object queries, DETR reasons about the relations of the objects and the global image context to directly output the final set of predictions in parallel. The new model is conceptually simple and does not require a specialized library, unlike many other modern detectors. DETR demonstrates accuracy and run-time performance on par with the well-established and highly-optimized Faster RCNN baseline on the challenging COCO object detection dataset. Moreover, DETR can be easily generalized to produce panoptic segmentation in a unified manner. We show that it significantly outperforms competitive baselines. Training code and pretrained models are available at this https URL.", "motivation": "We present a new method that views object detection as a direct set prediction problem.", "implementation": "The main ingredients of the new framework, called DEtection TRansformer or DETR, are a set-based global loss that forces unique predictions via bipartite matching, and a transformer encoder-decoder architecture.", "application": "Our approach streamlines the detection pipeline, effectively removing the need for many hand-designed components like a non-maximum suppression procedure or anchor generation that explicitly encode our prior knowledge about the task.", "problem": "We present a new method that views object detection as a direct set prediction problem.", "contribution": "The main ingredients of the new framework, called DEtection TRansformer or DETR, are a set-based global loss that forces unique predictions via bipartite matching, and a transformer encoder-decoder architecture.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "End-to-End Object Detection with Transformers 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 18,118；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "ECCV", "Datasets / Benchmarks" ], "code": { "available": false, "url": "" }, "url": "https://www.semanticscholar.org/paper/962dc29fdc3fbdc5930a10aba114050b82fe5a3e", "updated": "2026-05-20", "topCitedRank": 8, "citationCount": 18118, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/962dc29fdc3fbdc5930a10aba114050b82fe5a3e", "doi": "10.1007/978-3-030-58452-8_13", "arxiv": "2005.12872" }, { "title": "Xception: Deep Learning with Depthwise Separable Convolutions", "authors": "François Chollet", "venue": "CVPR", "year": 2016, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "High", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "We present an interpretation of Inception modules in convolutional neural networks as being an intermediate step in-between regular convolution and the depthwise separable convolution operation (a depthwise convolution followed by a pointwise convolution). In this light, a depthwise separable convolution can be understood as an Inception module with a maximally large number of towers. This observation leads us to propose a novel deep convolutional neural network architecture inspired by Inception, where Inception modules have been replaced with depthwise separable convolutions. We show that this architecture, dubbed Xception, slightly outperforms Inception V3 on the ImageNet dataset (which Inception V3 was designed for), and significantly outperforms Inception V3 on a larger image classification dataset comprising 350 million images and 17,000 classes. Since the Xception architecture has the same number of parameters as Inception V3, the performance gains are not due to increased capacity but rather to a more efficient use of model parameters.", "motivation": "We present an interpretation of Inception modules in convolutional neural networks as being an intermediate step in-between regular convolution and the depthwise separable convolution operation (a depthwise convolution followed by a pointwise convolution).", "implementation": "In this light, a depthwise separable convolution can be understood as an Inception module with a maximally large number of towers.", "application": "We show that this architecture, dubbed Xception, slightly outperforms Inception V3 on the ImageNet dataset (which Inception V3 was designed for), and significantly outperforms Inception V3 on a larger image classification dataset comprising 350 million images and 17,000 classes.", "problem": "We present an interpretation of Inception modules in convolutional neural networks as being an intermediate step in-between regular convolution and the depthwise separable convolution operation (a depthwise convolution followed by a pointwise convolution).", "contribution": "In this light, a depthwise separable convolution can be understood as an Inception module with a maximally large number of towers.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Xception: Deep Learning with Depthwise Separable Convolutions 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 17,642；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Datasets / Benchmarks" ], "code": { "available": false, "url": "" }, "url": "https://arxiv.org/pdf/1610.02357", "updated": "2026-05-20", "topCitedRank": 9, "citationCount": 17642, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/5b6ec746d309b165f9f9def873a2375b6fb40f3d", "doi": "10.1109/CVPR.2017.195", "arxiv": "1610.02357" }, { "title": "YOLO9000: Better, Faster, Stronger", "authors": "J. Redmon, Ali Farhadi", "venue": "CVPR", "year": 2016, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "High", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "We introduce YOLO9000, a state-of-the-art, real-time object detection system that can detect over 9000 object categories. First we propose various improvements to the YOLO detection method, both novel and drawn from prior work. The improved model, YOLOv2, is state-of-the-art on standard detection tasks like PASCAL VOC and COCO. Using a novel, multi-scale training method the same YOLOv2 model can run at varying sizes, offering an easy tradeoff between speed and accuracy. At 67 FPS, YOLOv2 gets 76.8 mAP on VOC 2007. At 40 FPS, YOLOv2 gets 78.6 mAP, outperforming state-of-the-art methods like Faster RCNN with ResNet and SSD while still running significantly faster. Finally we propose a method to jointly train on object detection and classification. Using this method we train YOLO9000 simultaneously on the COCO detection dataset and the ImageNet classification dataset. Our joint training allows YOLO9000 to predict detections for object classes that dont have labelled detection data. We validate our approach on the ImageNet detection task. YOLO9000 gets 19.7 mAP on the ImageNet detection validation set despite only having detection data for 44 of the 200 classes. On the 156 classes not in COCO, YOLO9000 gets 16.0 mAP. YOLO9000 predicts detections for more than 9000 different object categories, all in real-time.", "motivation": "YOLO9000 gets 19.7 mAP on the ImageNet detection validation set despite only having detection data for 44 of the 200 classes.", "implementation": "We introduce YOLO9000, a state-of-the-art, real-time object detection system that can detect over 9000 object categories.", "application": "First we propose various improvements to the YOLO detection method, both novel and drawn from prior work.", "problem": "YOLO9000 gets 19.7 mAP on the ImageNet detection validation set despite only having detection data for 44 of the 200 classes.", "contribution": "We introduce YOLO9000, a state-of-the-art, real-time object detection system that can detect over 9000 object categories.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "YOLO9000: Better, Faster, Stronger 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 17,517；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Datasets / Benchmarks", "award-winning", "Honorable Mention" ], "code": { "available": false, "url": "" }, "url": "https://arxiv.org/pdf/1612.08242", "updated": "2026-05-20", "topCitedRank": 10, "citationCount": 17517, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/7d39d69b23424446f0400ef603b2e3e22d0309d6", "doi": "10.1109/CVPR.2017.690", "arxiv": "1612.08242", "awards": [ { "conference": "CVPR", "year": 2017, "award": "Honorable Mention", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ] }, { "title": "PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation", "authors": "C. Qi, Hao Su, Kaichun Mo et al.", "venue": "CVPR", "year": 2016, "topic": "Segmentation", "modality": "Dense prediction", "status": "CV 顶会高引 Top 100", "priority": "High", "task": "语义分割、实例分割或全景分割", "method": "密集预测网络、上下文建模或统一分割框架", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "Point cloud is an important type of geometric data structure. Due to its irregular format, most researchers transform such data to regular 3D voxel grids or collections of images. This, however, renders data unnecessarily voluminous and causes issues. In this paper, we design a novel type of neural network that directly consumes point clouds, which well respects the permutation invariance of points in the input. Our network, named PointNet, provides a unified architecture for applications ranging from object classification, part segmentation, to scene semantic parsing. Though simple, PointNet is highly efficient and effective. Empirically, it shows strong performance on par or even better than state of the art. Theoretically, we provide analysis towards understanding of what the network has learnt and why the network is robust with respect to input perturbation and corruption.", "motivation": "This, however, renders data unnecessarily voluminous and causes issues.", "implementation": "In this paper, we design a novel type of neural network that directly consumes point clouds, which well respects the permutation invariance of points in the input.", "application": "Our network, named PointNet, provides a unified architecture for applications ranging from object classification, part segmentation, to scene semantic parsing.", "problem": "This, however, renders data unnecessarily voluminous and causes issues.", "contribution": "In this paper, we design a novel type of neural network that directly consumes point clouds, which well respects the permutation invariance of points in the input.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Segmentation 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 17,472；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何在高分辨率、开放类别和低标注条件下保持边界和语义一致。", "directionNote": "该方向的高引用工作集中体现了 Segmentation 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Segmentation" ], "code": { "available": false, "url": "" }, "url": "https://arxiv.org/pdf/1612.00593", "updated": "2026-05-20", "topCitedRank": 11, "citationCount": 17472, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/d997beefc0922d97202789d2ac307c55c2c52fba", "doi": "10.1109/CVPR.2017.16", "arxiv": "1612.00593" }, { "title": "Encoder-Decoder with Atrous Separable Convolution for Semantic Image Segmentation", "authors": "Liang-Chieh Chen, Yukun Zhu, G. Papandreou et al.", "venue": "ECCV", "year": 2018, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "High", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "Spatial pyramid pooling module or encode-decoder structure are used in deep neural networks for semantic segmentation task. The former networks are able to encode multi-scale contextual information by probing the incoming features with filters or pooling operations at multiple rates and multiple effective fields-of-view, while the latter networks can capture sharper object boundaries by gradually recovering the spatial information. In this work, we propose to combine the advantages from both methods. Specifically, our proposed model, DeepLabv3+, extends DeepLabv3 by adding a simple yet effective decoder module to refine the segmentation results especially along object boundaries. We further explore the Xception model and apply the depthwise separable convolution to both Atrous Spatial Pyramid Pooling and decoder modules, resulting in a faster and stronger encoder-decoder network. We demonstrate the effectiveness of the proposed model on PASCAL VOC 2012 and Cityscapes datasets, achieving the test set performance of 89.0\\% and 82.1\\% without any post-processing. Our paper is accompanied with a publicly available reference implementation of the proposed models in Tensorflow at \\url{this https URL}.", "motivation": "Spatial pyramid pooling module or encode-decoder structure are used in deep neural networks for semantic segmentation task.", "implementation": "In this work, we propose to combine the advantages from both methods.", "application": "Specifically, our proposed model, DeepLabv3+, extends DeepLabv3 by adding a simple yet effective decoder module to refine the segmentation results especially along object boundaries.", "problem": "Spatial pyramid pooling module or encode-decoder structure are used in deep neural networks for semantic segmentation task.", "contribution": "In this work, we propose to combine the advantages from both methods.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Encoder-Decoder with Atrous Separable Convolution for Semantic Image Segmentation 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 16,525；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "ECCV", "Datasets / Benchmarks" ], "code": { "available": false, "url": "" }, "url": "https://www.semanticscholar.org/paper/9217e28b2273eb3b26e4e9b7b498b4661e6e09f5", "updated": "2026-05-20", "topCitedRank": 12, "citationCount": 16525, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/9217e28b2273eb3b26e4e9b7b498b4661e6e09f5", "doi": "10.1007/978-3-030-01234-2_49", "arxiv": "1802.02611" }, { "title": "Momentum Contrast for Unsupervised Visual Representation Learning", "authors": "Kaiming He, Haoqi Fan, Yuxin Wu et al.", "venue": "CVPR", "year": 2019, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "High", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "We present Momentum Contrast (MoCo) for unsupervised visual representation learning. From a perspective on contrastive learning as dictionary look-up, we build a dynamic dictionary with a queue and a moving-averaged encoder. This enables building a large and consistent dictionary on-the-fly that facilitates contrastive unsupervised learning. MoCo provides competitive results under the common linear protocol on ImageNet classification. More importantly, the representations learned by MoCo transfer well to downstream tasks. MoCo can outperform its supervised pre-training counterpart in 7 detection/segmentation tasks on PASCAL VOC, COCO, and other datasets, sometimes surpassing it by large margins. This suggests that the gap between unsupervised and supervised representation learning has been largely closed in many vision tasks.", "motivation": "We present Momentum Contrast (MoCo) for unsupervised visual representation learning.", "implementation": "This enables building a large and consistent dictionary on-the-fly that facilitates contrastive unsupervised learning.", "application": "MoCo provides competitive results under the common linear protocol on ImageNet classification.", "problem": "We present Momentum Contrast (MoCo) for unsupervised visual representation learning.", "contribution": "This enables building a large and consistent dictionary on-the-fly that facilitates contrastive unsupervised learning.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Momentum Contrast for Unsupervised Visual Representation Learning 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 14,888；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Datasets / Benchmarks" ], "code": { "available": false, "url": "" }, "url": "https://arxiv.org/pdf/1911.05722", "updated": "2026-05-20", "topCitedRank": 13, "citationCount": 14888, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/add2f205338d70e10ce5e686df4a690e2851bdfc", "doi": "10.1109/cvpr42600.2020.00975", "arxiv": "1911.05722" }, { "title": "Pyramid Scene Parsing Network", "authors": "Hengshuang Zhao, Jianping Shi, Xiaojuan Qi et al.", "venue": "CVPR", "year": 2016, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "High", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "Scene parsing is challenging for unrestricted open vocabulary and diverse scenes. In this paper, we exploit the capability of global context information by different-region-based context aggregation through our pyramid pooling module together with the proposed pyramid scene parsing network (PSPNet). Our global prior representation is effective to produce good quality results on the scene parsing task, while PSPNet provides a superior framework for pixel-level prediction. The proposed approach achieves state-of-the-art performance on various datasets. It came first in ImageNet scene parsing challenge 2016, PASCAL VOC 2012 benchmark and Cityscapes benchmark. A single PSPNet yields the new record of mIoU accuracy 85.4% on PASCAL VOC 2012 and accuracy 80.2% on Cityscapes.", "motivation": "Scene parsing is challenging for unrestricted open vocabulary and diverse scenes.", "implementation": "In this paper, we exploit the capability of global context information by different-region-based context aggregation through our pyramid pooling module together with the proposed pyramid scene parsing network (PSPNet).", "application": "It came first in ImageNet scene parsing challenge 2016, PASCAL VOC 2012 benchmark and Cityscapes benchmark.", "problem": "Scene parsing is challenging for unrestricted open vocabulary and diverse scenes.", "contribution": "In this paper, we exploit the capability of global context information by different-region-based context aggregation through our pyramid pooling module together with the proposed pyramid scene parsing network (PSPNet).", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Pyramid Scene Parsing Network 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 14,090；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Datasets / Benchmarks" ], "code": { "available": false, "url": "" }, "url": "https://arxiv.org/pdf/1612.01105", "updated": "2026-05-20", "topCitedRank": 14, "citationCount": 14090, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/1031a69923b80ad01cf3fbb703d10757a80e699b", "doi": "10.1109/CVPR.2017.660", "arxiv": "1612.01105" }, { "title": "The Cityscapes Dataset for Semantic Urban Scene Understanding", "authors": "Marius Cordts, Mohamed Omran, Sebastian Ramos et al.", "venue": "CVPR", "year": 2016, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "High", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "Visual understanding of complex urban street scenes is an enabling factor for a wide range of applications. Object detection has benefited enormously from large-scale datasets, especially in the context of deep learning. For semantic urban scene understanding, however, no current dataset adequately captures the complexity of real-world urban scenes. To address this, we introduce Cityscapes, a benchmark suite and large-scale dataset to train and test approaches for pixel-level and instance-level semantic labeling. Cityscapes is comprised of a large, diverse set of stereo video sequences recorded in streets from 50 different cities. 5000 of these images have high quality pixel-level annotations, 20 000 additional images have coarse annotations to enable methods that leverage large volumes of weakly-labeled data. Crucially, our effort exceeds previous attempts in terms of dataset size, annotation richness, scene variability, and complexity. Our accompanying empirical study provides an in-depth analysis of the dataset characteristics, as well as a performance evaluation of several state-of-the-art approaches based on our benchmark.", "motivation": "For semantic urban scene understanding, however, no current dataset adequately captures the complexity of real-world urban scenes.", "implementation": "To address this, we introduce Cityscapes, a benchmark suite and large-scale dataset to train and test approaches for pixel-level and instance-level semantic labeling.", "application": "Visual understanding of complex urban street scenes is an enabling factor for a wide range of applications.", "problem": "For semantic urban scene understanding, however, no current dataset adequately captures the complexity of real-world urban scenes.", "contribution": "To address this, we introduce Cityscapes, a benchmark suite and large-scale dataset to train and test approaches for pixel-level and instance-level semantic labeling.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "The Cityscapes Dataset for Semantic Urban Scene Understanding 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 13,474；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Datasets / Benchmarks" ], "code": { "available": false, "url": "" }, "url": "https://arxiv.org/pdf/1604.01685", "updated": "2026-05-20", "topCitedRank": 15, "citationCount": 13474, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/c8c494ee5488fe20e0aa01bddf3fc4632086d654", "doi": "10.1109/CVPR.2016.350", "arxiv": "1604.01685" }, { "title": "Segment Anything", "authors": "A. Kirillov, Eric Mintun, Nikhila Ravi et al.", "venue": "ICCV", "year": 2023, "topic": "Vision-Language / Foundation Models", "modality": "Image-text / foundation model", "status": "CV 顶会高引 Top 100", "priority": "High", "task": "视觉-语言理解、开放词表识别或基础模型适配", "method": "大规模预训练、多模态对齐或提示式视觉建模", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "We introduce the Segment Anything (SA) project: a new task, model, and dataset for image segmentation. Using our efficient model in a data collection loop, we built the largest segmentation dataset to date (by far), with over 1 billion masks on 11M licensed and privacy respecting images. The model is designed and trained to be promptable, so it can transfer zero-shot to new image distributions and tasks. We evaluate its capabilities on numerous tasks and find that its zero-shot performance is impressive – often competitive with or even superior to prior fully supervised results. We are releasing the Segment Anything Model (SAM) and corresponding dataset (SA-1B) of 1B masks and 11M images at segment-anything.com to foster research into foundation models for computer vision. We recommend reading the full paper at: arxiv.org/abs/2304.02643.", "motivation": "We introduce the Segment Anything (SA) project: a new task, model, and dataset for image segmentation.", "implementation": "Using our efficient model in a data collection loop, we built the largest segmentation dataset to date (by far), with over 1 billion masks on 11M licensed and privacy respecting images.", "application": "We evaluate its capabilities on numerous tasks and find that its zero-shot performance is impressive – often competitive with or even superior to prior fully supervised results.", "problem": "We introduce the Segment Anything (SA) project: a new task, model, and dataset for image segmentation.", "contribution": "Using our efficient model in a data collection loop, we built the largest segmentation dataset to date (by far), with over 1 billion masks on 11M licensed and privacy respecting images.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Segment Anything 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Vision-Language / Foundation Models 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 13,432；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何降低标注和算力成本，同时保持可控性、可解释性和领域泛化。", "directionNote": "该方向的高引用工作集中体现了 Vision-Language / Foundation Models 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "ICCV", "Vision-Language / Foundation Models", "award-winning", "Honorable Mention" ], "code": { "available": false, "url": "" }, "url": "https://www.semanticscholar.org/paper/7470a1702c8c86e6f28d32cfa315381150102f5b", "updated": "2026-05-20", "topCitedRank": 16, "citationCount": 13432, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/7470a1702c8c86e6f28d32cfa315381150102f5b", "doi": "10.1109/ICCV51070.2023.00371", "arxiv": "2304.02643", "awards": [ { "conference": "ICCV", "year": 2023, "award": "Honorable Mention", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ] }, { "title": "A Style-Based Generator Architecture for Generative Adversarial Networks", "authors": "Tero Karras, S. Laine, Timo Aila", "venue": "CVPR", "year": 2018, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "High", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "We propose an alternative generator architecture for generative adversarial networks, borrowing from style transfer literature. The new architecture leads to an automatically learned, unsupervised separation of high-level attributes (e.g., pose and identity when trained on human faces) and stochastic variation in the generated images (e.g., freckles, hair), and it enables intuitive, scale-specific control of the synthesis. The new generator improves the state-of-the-art in terms of traditional distribution quality metrics, leads to demonstrably better interpolation properties, and also better disentangles the latent factors of variation. To quantify interpolation quality and disentanglement, we propose two new, automated methods that are applicable to any generator architecture. Finally, we introduce a new, highly varied and high-quality dataset of human faces.", "motivation": "We propose an alternative generator architecture for generative adversarial networks, borrowing from style transfer literature.", "implementation": "The new architecture leads to an automatically learned, unsupervised separation of high-level attributes (e.g., pose and identity when trained on human faces) and stochastic variation in the generated images (e.g., freckles, hair), and it enables intuitive, scale-specific control of the synthesis.", "application": "Finally, we introduce a new, highly varied and high-quality dataset of human faces.", "problem": "We propose an alternative generator architecture for generative adversarial networks, borrowing from style transfer literature.", "contribution": "The new architecture leads to an automatically learned, unsupervised separation of high-level attributes (e.g., pose and identity when trained on human faces) and stochastic variation in the generated images (e.g., freckles, hair), and it enables intuitive, scale-specific control of the synthesis.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "A Style-Based Generator Architecture for Generative Adversarial Networks 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 12,885；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Datasets / Benchmarks", "award-winning", "Honorable Mention" ], "code": { "available": false, "url": "" }, "url": "https://arxiv.org/pdf/1812.04948", "updated": "2026-05-20", "topCitedRank": 17, "citationCount": 12885, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/ceb2ebef0b41e31c1a21b28c2734123900c005e2", "doi": "10.1109/CVPR.2019.00453", "arxiv": "1812.04948", "awards": [ { "conference": "CVPR", "year": 2019, "award": "Honorable Mention", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ] }, { "title": "Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network", "authors": "C. Ledig, Lucas Theis, Ferenc Huszár et al.", "venue": "CVPR", "year": 2016, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "High", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "Despite the breakthroughs in accuracy and speed of single image super-resolution using faster and deeper convolutional neural networks, one central problem remains largely unsolved: how do we recover the finer texture details when we super-resolve at large upscaling factors? The behavior of optimization-based super-resolution methods is principally driven by the choice of the objective function. Recent work has largely focused on minimizing the mean squared reconstruction error. The resulting estimates have high peak signal-to-noise ratios, but they are often lacking high-frequency details and are perceptually unsatisfying in the sense that they fail to match the fidelity expected at the higher resolution. In this paper, we present SRGAN, a generative adversarial network (GAN) for image super-resolution (SR). To our knowledge, it is the first framework capable of inferring photo-realistic natural images for 4x upscaling factors. To achieve this, we propose a perceptual loss function which consists of an adversarial loss and a content loss. The adversarial loss pushes our solution to the natural image manifold using a discriminator network that is trained to differentiate between the super-resolved images and original photo-realistic images. In addition, we use a content loss motivated by perceptual similarity instead of similarity in pixel space. Our deep residual network is able to recover photo-realistic textures from heavily downsampled images on public benchmarks. An extensive mean-opinion-score (MOS) test shows hugely significant gains in perceptual quality using SRGAN. The MOS scores obtained with SRGAN are closer to those of the original high-resolution images than to those obtained with any state-of-the-art method.", "motivation": "Despite the breakthroughs in accuracy and speed of single image super-resolution using faster and deeper convolutional neural networks, one central problem remains largely unsolved: how do we recover the finer texture details when we super-resolve at large upscaling factors?", "implementation": "In this paper, we present SRGAN, a generative adversarial network (GAN) for image super-resolution (SR).", "application": "Recent work has largely focused on minimizing the mean squared reconstruction error.", "problem": "Despite the breakthroughs in accuracy and speed of single image super-resolution using faster and deeper convolutional neural networks, one central problem remains largely unsolved: how do we recover the finer texture details when we super-resolve at large upscaling factors?", "contribution": "In this paper, we present SRGAN, a generative adversarial network (GAN) for image super-resolution (SR).", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 12,049；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Datasets / Benchmarks" ], "code": { "available": false, "url": "" }, "url": "http://arxiv.org/pdf/1609.04802", "updated": "2026-05-20", "topCitedRank": 18, "citationCount": 12049, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/df0c54fe61f0ffb9f0e36a17c2038d9a1964cba3", "doi": "10.1109/CVPR.2017.19", "arxiv": "1609.04802" }, { "title": "Aggregated Residual Transformations for Deep Neural Networks", "authors": "Saining Xie, Ross B. Girshick, Piotr Dollár et al.", "venue": "CVPR", "year": 2016, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "High", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "We present a simple, highly modularized network architecture for image classification. Our network is constructed by repeating a building block that aggregates a set of transformations with the same topology. Our simple design results in a homogeneous, multi-branch architecture that has only a few hyper-parameters to set. This strategy exposes a new dimension, which we call cardinality (the size of the set of transformations), as an essential factor in addition to the dimensions of depth and width. On the ImageNet-1K dataset, we empirically show that even under the restricted condition of maintaining complexity, increasing cardinality is able to improve classification accuracy. Moreover, increasing cardinality is more effective than going deeper or wider when we increase the capacity. Our models, named ResNeXt, are the foundations of our entry to the ILSVRC 2016 classification task in which we secured 2nd place. We further investigate ResNeXt on an ImageNet-5K set and the COCO detection set, also showing better results than its ResNet counterpart. The code and models are publicly available online.", "motivation": "We present a simple, highly modularized network architecture for image classification.", "implementation": "Our network is constructed by repeating a building block that aggregates a set of transformations with the same topology.", "application": "On the ImageNet-1K dataset, we empirically show that even under the restricted condition of maintaining complexity, increasing cardinality is able to improve classification accuracy.", "problem": "We present a simple, highly modularized network architecture for image classification.", "contribution": "Our network is constructed by repeating a building block that aggregates a set of transformations with the same topology.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Aggregated Residual Transformations for Deep Neural Networks 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 11,614；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Datasets / Benchmarks" ], "code": { "available": false, "url": "" }, "url": "http://arxiv.org/pdf/1611.05431", "updated": "2026-05-20", "topCitedRank": 19, "citationCount": 11614, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/f6e0856b4a9199fa968ac00da612a9407b5cb85c", "doi": "10.1109/CVPR.2017.634", "arxiv": "1611.05431" }, { "title": "Masked Autoencoders Are Scalable Vision Learners", "authors": "Kaiming He, Xinlei Chen, Saining Xie et al.", "venue": "CVPR", "year": 2021, "topic": "Representation Learning", "modality": "Visual representation", "status": "CV 顶会高引 Top 100", "priority": "High", "task": "自监督、半监督或可迁移视觉表征学习", "method": "对比学习、掩码建模、数据增强或持续学习", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "This paper shows that masked autoencoders (MAE) are scalable self-supervised learners for computer vision. Our MAE approach is simple: we mask random patches of the input image and reconstruct the missing pixels. It is based on two core designs. First, we develop an asymmetric encoder-decoder architecture, with an encoder that operates only on the visible subset of patches (without mask tokens), along with a lightweight decoder that reconstructs the original image from the latent representation and mask tokens. Second, we find that masking a high proportion of the input image, e.g., 75%, yields a nontrivial and meaningful self-supervisory task. Coupling these two designs enables us to train large models efficiently and effectively: we accelerate training (by 3× or more) and improve accuracy. Our scalable approach allows for learning high-capacity models that generalize well: e.g., a vanilla ViT-Huge model achieves the best accuracy (87.8%) among methods that use only ImageNet-1K data. Transfer performance in downstream tasks outperforms supervised pretraining and shows promising scaling behavior.", "motivation": "This paper shows that masked autoencoders (MAE) are scalable self-supervised learners for computer vision.", "implementation": "First, we develop an asymmetric encoder-decoder architecture, with an encoder that operates only on the visible subset of patches (without mask tokens), along with a lightweight decoder that reconstructs the original image from the latent representation and mask tokens.", "application": "Our scalable approach allows for learning high-capacity models that generalize well: e.g., a vanilla ViT-Huge model achieves the best accuracy (87.8%) among methods that use only ImageNet-1K data.", "problem": "This paper shows that masked autoencoders (MAE) are scalable self-supervised learners for computer vision.", "contribution": "First, we develop an asymmetric encoder-decoder architecture, with an encoder that operates only on the visible subset of patches (without mask tokens), along with a lightweight decoder that reconstructs the original image from the latent representation and mask tokens.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Masked Autoencoders Are Scalable Vision Learners 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Representation Learning 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 11,418；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何评估表征的因果性、可迁移性和下游任务稳定性。", "directionNote": "该方向的高引用工作集中体现了 Representation Learning 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Representation Learning" ], "code": { "available": false, "url": "" }, "url": "https://arxiv.org/pdf/2111.06377", "updated": "2026-05-20", "topCitedRank": 20, "citationCount": 11418, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/6351ebb4a3287f5f3e1273464b3b91e5df5a16d7", "doi": "10.1109/CVPR52688.2022.01553", "arxiv": "2111.06377" }, { "title": "Perceptual Losses for Real-Time Style Transfer and Super-Resolution", "authors": "Justin Johnson, Alexandre Alahi, Li Fei-Fei", "venue": "ECCV", "year": 2016, "topic": "Generative Vision", "modality": "Image generation / editing", "status": "CV 顶会高引 Top 100", "priority": "High", "task": "图像生成、编辑、翻译或高保真合成", "method": "GAN、diffusion、transformer 或条件生成模型", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "We consider image transformation problems, where an input image is transformed into an output image. Recent methods for such problems typically train feed-forward convolutional neural networks using a per-pixel loss between the output and ground-truth images. Parallel work has shown that high-quality images can be generated by defining and optimizing perceptual loss functions based on high-level features extracted from pretrained networks. We combine the benefits of both approaches, and propose the use of perceptual loss functions for training feed-forward networks for image transformation tasks. We show results on image style transfer, where a feed-forward network is trained to solve the optimization problem proposed by Gatys et al. in real-time. Compared to the optimization-based method, our network gives similar qualitative results but is three orders of magnitude faster. We also experiment with single-image super-resolution, where replacing a per-pixel loss with a perceptual loss gives visually pleasing results.", "motivation": "We show results on image style transfer, where a feed-forward network is trained to solve the optimization problem proposed by Gatys et al.", "implementation": "Recent methods for such problems typically train feed-forward convolutional neural networks using a per-pixel loss between the output and ground-truth images.", "application": "We combine the benefits of both approaches, and propose the use of perceptual loss functions for training feed-forward networks for image transformation tasks.", "problem": "We show results on image style transfer, where a feed-forward network is trained to solve the optimization problem proposed by Gatys et al.", "contribution": "Recent methods for such problems typically train feed-forward convolutional neural networks using a per-pixel loss between the output and ground-truth images.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Perceptual Losses for Real-Time Style Transfer and Super-Resolution 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Generative Vision 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 11,410；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何平衡生成质量、可控性、版权风险和真实任务可靠性。", "directionNote": "该方向的高引用工作集中体现了 Generative Vision 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "ECCV", "Generative Vision" ], "code": { "available": false, "url": "" }, "url": "https://www.semanticscholar.org/paper/915c4bb289b3642489e904c65a47fa56efb60658", "updated": "2026-05-20", "topCitedRank": 21, "citationCount": 11410, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/915c4bb289b3642489e904c65a47fa56efb60658", "doi": "10.1007/978-3-319-46475-6_43", "arxiv": "1603.08155" }, { "title": "Identity Mappings in Deep Residual Networks", "authors": "Kaiming He, X. Zhang, Shaoqing Ren et al.", "venue": "ECCV", "year": 2016, "topic": "Video / Human Understanding", "modality": "Video / human-centric data", "status": "CV 顶会高引 Top 100", "priority": "High", "task": "视频理解、动作识别、人体姿态或轨迹预测", "method": "时序建模、人体结构先验或视频 transformer", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "Deep residual networks have emerged as a family of extremely deep architectures showing compelling accuracy and nice convergence behaviors. In this paper, we analyze the propagation formulations behind the residual building blocks, which suggest that the forward and backward signals can be directly propagated from one block to any other block, when using identity mappings as the skip connections and after-addition activation. A series of ablation experiments support the importance of these identity mappings. This motivates us to propose a new residual unit, which makes training easier and improves generalization. We report improved results using a 1001-layer ResNet on CIFAR-10 (4.62 % error) and CIFAR-100, and a 200-layer ResNet on ImageNet. Code is available at: https://github.com/KaimingHe/resnet-1k-layers.", "motivation": "Deep residual networks have emerged as a family of extremely deep architectures showing compelling accuracy and nice convergence behaviors.", "implementation": "A series of ablation experiments support the importance of these identity mappings.", "application": "We report improved results using a 1001-layer ResNet on CIFAR-10 (4.62 % error) and CIFAR-100, and a 200-layer ResNet on ImageNet.", "problem": "Deep residual networks have emerged as a family of extremely deep architectures showing compelling accuracy and nice convergence behaviors.", "contribution": "A series of ablation experiments support the importance of these identity mappings.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Identity Mappings in Deep Residual Networks 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Video / Human Understanding 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 11,153；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何在长时序、遮挡和复杂交互下保持鲁棒理解。", "directionNote": "该方向的高引用工作集中体现了 Video / Human Understanding 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "ECCV", "Video / Human Understanding" ], "code": { "available": true, "url": "https://github.com/KaimingHe/resnet-1k-layers" }, "url": "https://www.semanticscholar.org/paper/77f0a39b8e02686fd85b01971f8feb7f60971f80", "updated": "2026-05-20", "topCitedRank": 22, "citationCount": 11153, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/77f0a39b8e02686fd85b01971f8feb7f60971f80", "doi": "10.1007/978-3-319-46493-0_38", "arxiv": "1603.05027" }, { "title": "YOLOv7: Trainable Bag-of-Freebies Sets New State-of-the-Art for Real-Time Object Detectors", "authors": "Chien-Yao Wang, Alexey Bochkovskiy, H. Liao", "venue": "CVPR", "year": 2022, "topic": "Object Detection", "modality": "RGB / multi-scale visual features", "status": "CV 顶会高引 Top 100", "priority": "High", "task": "目标检测、开放集检测或检测损失设计", "method": "检测框架、特征金字塔、transformer detector 或检测损失", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "Real-time object detection is one of the most important research topics in computer vision. As new approaches regarding architecture optimization and training optimization are continually being developed, we have found two research topics that have spawned when dealing with these latest state-of-the-art methods. To address the topics, we propose a trainable bag-of-freebies oriented solution. We combine the flexible and efficient training tools with the proposed architecture and the compound scaling method. YOLOv7 surpasses all known object detectors in both speed and accuracy in the range from 5 FPS to 120 FPS and has the highest accuracy 56.8% AP among all known real-time object detectors with 30 FPS or higher on GPU V100. Source code is released in https://github.com/WongKinYiu/yolov7.", "motivation": "Real-time object detection is one of the most important research topics in computer vision.", "implementation": "As new approaches regarding architecture optimization and training optimization are continually being developed, we have found two research topics that have spawned when dealing with these latest state-of-the-art methods.", "application": "YOLOv7 surpasses all known object detectors in both speed and accuracy in the range from 5 FPS to 120 FPS and has the highest accuracy 56.8% AP among all known real-time object detectors with 30 FPS or higher on GPU V100.", "problem": "Real-time object detection is one of the most important research topics in computer vision.", "contribution": "As new approaches regarding architecture optimization and training optimization are continually being developed, we have found two research topics that have spawned when dealing with these latest state-of-the-art methods.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "YOLOv7: Trainable Bag-of-Freebies Sets New State-of-the-Art for Real-Time Object Detectors 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Object Detection 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 10,170；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何在开放世界、长尾类别和实时部署中保持稳定检测性能。", "directionNote": "该方向的高引用工作集中体现了 Object Detection 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Object Detection" ], "code": { "available": true, "url": "https://github.com/WongKinYiu/yolov7" }, "url": "https://arxiv.org/pdf/2207.02696", "updated": "2026-05-20", "topCitedRank": 23, "citationCount": 10170, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/3aed4648f7857c1d5e9b1da4c3afaf97463138c3", "doi": "10.1109/CVPR52729.2023.00721", "arxiv": "2207.02696" }, { "title": "Quo Vadis, Action Recognition? A New Model and the Kinetics Dataset", "authors": "João Carreira, Andrew Zisserman", "venue": "CVPR", "year": 2017, "topic": "Vision-Language / Foundation Models", "modality": "Image-text / foundation model", "status": "CV 顶会高引 Top 100", "priority": "High", "task": "视觉-语言理解、开放词表识别或基础模型适配", "method": "大规模预训练、多模态对齐或提示式视觉建模", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "The paucity of videos in current action classification datasets (UCF-101 and HMDB-51) has made it difficult to identify good video architectures, as most methods obtain similar performance on existing small-scale benchmarks. This paper re-evaluates state-of-the-art architectures in light of the new Kinetics Human Action Video dataset. Kinetics has two orders of magnitude more data, with 400 human action classes and over 400 clips per class, and is collected from realistic, challenging YouTube videos. We provide an analysis on how current architectures fare on the task of action classification on this dataset and how much performance improves on the smaller benchmark datasets after pre-training on Kinetics. We also introduce a new Two-Stream Inflated 3D ConvNet (I3D) that is based on 2D ConvNet inflation: filters and pooling kernels of very deep image classification ConvNets are expanded into 3D, making it possible to learn seamless spatio-temporal feature extractors from video while leveraging successful ImageNet architecture designs and even their parameters. We show that, after pre-training on Kinetics, I3D models considerably improve upon the state-of-the-art in action classification, reaching 80.2% on HMDB-51 and 97.9% on UCF-101.", "motivation": "The paucity of videos in current action classification datasets (UCF-101 and HMDB-51) has made it difficult to identify good video architectures, as most methods obtain similar performance on existing small-scale benchmarks.", "implementation": "This paper re-evaluates state-of-the-art architectures in light of the new Kinetics Human Action Video dataset.", "application": "We provide an analysis on how current architectures fare on the task of action classification on this dataset and how much performance improves on the smaller benchmark datasets after pre-training on Kinetics.", "problem": "The paucity of videos in current action classification datasets (UCF-101 and HMDB-51) has made it difficult to identify good video architectures, as most methods obtain similar performance on existing small-scale benchmarks.", "contribution": "This paper re-evaluates state-of-the-art architectures in light of the new Kinetics Human Action Video dataset.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Quo Vadis, Action Recognition? A New Model and the Kinetics Dataset 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Vision-Language / Foundation Models 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 9,488；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何降低标注和算力成本，同时保持可控性、可解释性和领域泛化。", "directionNote": "该方向的高引用工作集中体现了 Vision-Language / Foundation Models 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Vision-Language / Foundation Models" ], "code": { "available": false, "url": "" }, "url": "http://arxiv.org/pdf/1705.07750", "updated": "2026-05-20", "topCitedRank": 24, "citationCount": 9488, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/b61a3f8b80bbd44f24544dc915f52fd30bbdf485", "doi": "10.1109/CVPR.2017.502", "arxiv": "1705.07750" }, { "title": "Searching for MobileNetV3", "authors": "Andrew G. Howard, M. Sandler, Grace Chu et al.", "venue": "ICCV", "year": 2019, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "High", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "We present the next generation of MobileNets based on a combination of complementary search techniques as well as a novel architecture design. MobileNetV3 is tuned to mobile phone CPUs through a combination of hardware-aware network architecture search (NAS) complemented by the NetAdapt algorithm and then subsequently improved through novel architecture advances. This paper starts the exploration of how automated search algorithms and network design can work together to harness complementary approaches improving the overall state of the art. Through this process we create two new MobileNet models for release: MobileNetV3-Large and MobileNetV3-Small which are targeted for high and low resource use cases. These models are then adapted and applied to the tasks of object detection and semantic segmentation. For the task of semantic segmentation (or any dense pixel prediction), we propose a new efficient segmentation decoder Lite Reduced Atrous Spatial Pyramid Pooling (LR-ASPP). We achieve new state of the art results for mobile classification, detection and segmentation. MobileNetV3-Large is 3.2% more accurate on ImageNet classification while reducing latency by 20% compared to MobileNetV2. MobileNetV3-Small is 6.6% more accurate compared to a MobileNetV2 model with comparable latency. MobileNetV3-Large detection is over 25% faster at roughly the same accuracy as MobileNetV2 on COCO detection. MobileNetV3-Large LR-ASPP is 34% faster than MobileNetV2 R-ASPP at similar accuracy for Cityscapes segmentation.", "motivation": "We present the next generation of MobileNets based on a combination of complementary search techniques as well as a novel architecture design.", "implementation": "MobileNetV3 is tuned to mobile phone CPUs through a combination of hardware-aware network architecture search (NAS) complemented by the NetAdapt algorithm and then subsequently improved through novel architecture advances.", "application": "These models are then adapted and applied to the tasks of object detection and semantic segmentation.", "problem": "We present the next generation of MobileNets based on a combination of complementary search techniques as well as a novel architecture design.", "contribution": "MobileNetV3 is tuned to mobile phone CPUs through a combination of hardware-aware network architecture search (NAS) complemented by the NetAdapt algorithm and then subsequently improved through novel architecture advances.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Searching for MobileNetV3 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 9,245；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "ICCV", "Datasets / Benchmarks" ], "code": { "available": false, "url": "" }, "url": "https://www.semanticscholar.org/paper/5e19eba1e6644f7c83f607383d256deea71f87ae", "updated": "2026-05-20", "topCitedRank": 25, "citationCount": 9245, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/5e19eba1e6644f7c83f607383d256deea71f87ae", "doi": "10.1109/ICCV.2019.00140", "arxiv": "1905.02244" }, { "title": "Emerging Properties in Self-Supervised Vision Transformers", "authors": "Mathilde Caron, Hugo Touvron, Ishan Misra et al.", "venue": "ICCV", "year": 2021, "topic": "Segmentation", "modality": "Dense prediction", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "语义分割、实例分割或全景分割", "method": "密集预测网络、上下文建模或统一分割框架", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "In this paper, we question if self-supervised learning provides new properties to Vision Transformer (ViT) [16] that stand out compared to convolutional networks (convnets). Beyond the fact that adapting self-supervised methods to this architecture works particularly well, we make the following observations: first, self-supervised ViT features contain explicit information about the semantic segmentation of an image, which does not emerge as clearly with supervised ViTs, nor with convnets. Second, these features are also excellent k-NN classifiers, reaching 78.3% top-1 on ImageNet with a small ViT. Our study also underlines the importance of momentum encoder [26], multi-crop training [9], and the use of small patches with ViTs. We implement our findings into a simple self-supervised method, called DINO, which we interpret as a form of self-distillation with no labels. We show the synergy between DINO and ViTs by achieving 80.1% top-1 on ImageNet in linear evaluation with ViT-Base.", "motivation": "In this paper, we question if self-supervised learning provides new properties to Vision Transformer (ViT) [16] that stand out compared to convolutional networks (convnets).", "implementation": "Beyond the fact that adapting self-supervised methods to this architecture works particularly well, we make the following observations: first, self-supervised ViT features contain explicit information about the semantic segmentation of an image, which does not emerge as clearly with supervised ViTs, nor with convnets.", "application": "Second, these features are also excellent k-NN classifiers, reaching 78.3% top-1 on ImageNet with a small ViT.", "problem": "In this paper, we question if self-supervised learning provides new properties to Vision Transformer (ViT) [16] that stand out compared to convolutional networks (convnets).", "contribution": "Beyond the fact that adapting self-supervised methods to this architecture works particularly well, we make the following observations: first, self-supervised ViT features contain explicit information about the semantic segmentation of an image, which does not emerge as clearly with supervised ViTs, nor with convnets.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Emerging Properties in Self-Supervised Vision Transformers 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Segmentation 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 9,014；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何在高分辨率、开放类别和低标注条件下保持边界和语义一致。", "directionNote": "该方向的高引用工作集中体现了 Segmentation 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "ICCV", "Segmentation" ], "code": { "available": false, "url": "" }, "url": "http://arxiv.org/pdf/2104.14294", "updated": "2026-05-20", "topCitedRank": 26, "citationCount": 9014, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/ad4a0938c48e61b7827869e4ac3baffd0aefab35", "doi": "10.1109/ICCV48922.2021.00951", "arxiv": "2104.14294" }, { "title": "A ConvNet for the 2020s", "authors": "Zhuang Liu, Hanzi Mao, Chaozheng Wu et al.", "venue": "CVPR", "year": 2022, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "The “Roaring 20s” of visual recognition began with the introduction of Vision Transformers (ViTs), which quickly superseded ConvNets as the state-of-the-art image classification model. A vanilla ViT, on the other hand, faces difficulties when applied to general computer vision tasks such as object detection and semantic segmentation. It is the hierarchical Transformers (e.g., Swin Transformers) that reintroduced several ConvNet priors, making Transformers practically viable as a generic vision backbone and demonstrating remarkable performance on a wide variety of vision tasks. However, the effectiveness of such hybrid approaches is still largely credited to the intrinsic superiority of Transformers, rather than the inherent inductive biases of convolutions. In this work, we reexamine the design spaces and test the limits of what a pure ConvNet can achieve. We gradually “modernize” a standard ResNet toward the design of a vision Transformer, and discover several key components that contribute to the performance difference along the way. The outcome of this exploration is a family of pure ConvNet models dubbed ConvNeXt. Constructed entirely from standard ConvNet modules, ConvNeXts compete favorably with Transformers in terms of accuracy and scalability, achieving 87.8% ImageNet top-1 accuracy and outperforming Swin Transformers on COCO detection and ADE20K segmentation, while maintaining the simplicity and efficiency of standard ConvNets.", "motivation": "However, the effectiveness of such hybrid approaches is still largely credited to the intrinsic superiority of Transformers, rather than the inherent inductive biases of convolutions.", "implementation": "The “Roaring 20s” of visual recognition began with the introduction of Vision Transformers (ViTs), which quickly superseded ConvNets as the state-of-the-art image classification model.", "application": "A vanilla ViT, on the other hand, faces difficulties when applied to general computer vision tasks such as object detection and semantic segmentation.", "problem": "However, the effectiveness of such hybrid approaches is still largely credited to the intrinsic superiority of Transformers, rather than the inherent inductive biases of convolutions.", "contribution": "The “Roaring 20s” of visual recognition began with the introduction of Vision Transformers (ViTs), which quickly superseded ConvNets as the state-of-the-art image classification model.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "A ConvNet for the 2020s 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 8,117；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Datasets / Benchmarks" ], "code": { "available": false, "url": "" }, "url": "https://www.semanticscholar.org/paper/177e957f5cd93229c9794ea652c646d2557b4a69", "updated": "2026-05-20", "topCitedRank": 27, "citationCount": 8117, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/177e957f5cd93229c9794ea652c646d2557b4a69", "doi": "10.1109/CVPR52688.2022.01167", "arxiv": "2201.03545" }, { "title": "nuScenes: A Multimodal Dataset for Autonomous Driving", "authors": "Holger Caesar, Varun Bankiti, Alex H. Lang et al.", "venue": "CVPR", "year": 2019, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "Robust detection and tracking of objects is crucial for the deployment of autonomous vehicle technology. Image based benchmark datasets have driven development in computer vision tasks such as object detection, tracking and segmentation of agents in the environment. Most autonomous vehicles, however, carry a combination of cameras and range sensors such as lidar and radar. As machine learning based methods for detection and tracking become more prevalent, there is a need to train and evaluate such methods on datasets containing range sensor data along with images. In this work we present nuTonomy scenes (nuScenes), the first dataset to carry the full autonomous vehicle sensor suite: 6 cameras, 5 radars and 1 lidar, all with full 360 degree field of view. nuScenes comprises 1000 scenes, each 20s long and fully annotated with 3D bounding boxes for 23 classes and 8 attributes. It has 7x as many annotations and 100x as many images as the pioneering KITTI dataset. We define novel 3D detection and tracking metrics. We also provide careful dataset analysis as well as baselines for lidar and image based detection and tracking. Data, development kit and more information are available online.", "motivation": "Most autonomous vehicles, however, carry a combination of cameras and range sensors such as lidar and radar.", "implementation": "In this work we present nuTonomy scenes (nuScenes), the first dataset to carry the full autonomous vehicle sensor suite: 6 cameras, 5 radars and 1 lidar, all with full 360 degree field of view.", "application": "Robust detection and tracking of objects is crucial for the deployment of autonomous vehicle technology.", "problem": "Most autonomous vehicles, however, carry a combination of cameras and range sensors such as lidar and radar.", "contribution": "In this work we present nuTonomy scenes (nuScenes), the first dataset to carry the full autonomous vehicle sensor suite: 6 cameras, 5 radars and 1 lidar, all with full 360 degree field of view.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "nuScenes: A Multimodal Dataset for Autonomous Driving 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 8,063；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Datasets / Benchmarks" ], "code": { "available": false, "url": "" }, "url": "https://arxiv.org/pdf/1903.11027", "updated": "2026-05-20", "topCitedRank": 28, "citationCount": 8063, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/9e475a514f54665478aac6038c262e5a6bac5e64", "doi": "10.1109/cvpr42600.2020.01164", "arxiv": "1903.11027" }, { "title": "ArcFace: Additive Angular Margin Loss for Deep Face Recognition", "authors": "Jiankang Deng, J. Guo, S. Zafeiriou", "venue": "CVPR", "year": 2018, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "One of the main challenges in feature learning using Deep Convolutional Neural Networks (DCNNs) for large-scale face recognition is the design of appropriate loss functions that can enhance the discriminative power. Centre loss penalises the distance between deep features and their corresponding class centres in the Euclidean space to achieve intra-class compactness. SphereFace assumes that the linear transformation matrix in the last fully connected layer can be used as a representation of the class centres in the angular space and therefore penalises the angles between deep features and their corresponding weights in a multiplicative way. Recently, a popular line of research is to incorporate margins in well-established loss functions in order to maximise face class separability. In this paper, we propose an Additive Angular Margin Loss (ArcFace) to obtain highly discriminative features for face recognition. The proposed ArcFace has a clear geometric interpretation due to its exact correspondence to geodesic distance on a hypersphere. We present arguably the most extensive experimental evaluation against all recent state-of-the-art face recognition methods on ten face recognition benchmarks which includes a new large-scale image database with trillions of pairs and a large-scale video dataset. We show that ArcFace consistently outperforms the state of the art and can be easily implemented with negligible computational overhead. To facilitate future research, the code has been made available.", "motivation": "One of the main challenges in feature learning using Deep Convolutional Neural Networks (DCNNs) for large-scale face recognition is the design of appropriate loss functions that can enhance the discriminative power.", "implementation": "Centre loss penalises the distance between deep features and their corresponding class centres in the Euclidean space to achieve intra-class compactness.", "application": "SphereFace assumes that the linear transformation matrix in the last fully connected layer can be used as a representation of the class centres in the angular space and therefore penalises the angles between deep features and their corresponding weights in a multiplicative way.", "problem": "One of the main challenges in feature learning using Deep Convolutional Neural Networks (DCNNs) for large-scale face recognition is the design of appropriate loss functions that can enhance the discriminative power.", "contribution": "Centre loss penalises the distance between deep features and their corresponding class centres in the Euclidean space to achieve intra-class compactness.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "ArcFace: Additive Angular Margin Loss for Deep Face Recognition 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 7,721；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Datasets / Benchmarks" ], "code": { "available": false, "url": "" }, "url": "http://spiral.imperial.ac.uk/bitstream/10044/1/69953/2/1801.07698.pdf", "updated": "2026-05-20", "topCitedRank": 29, "citationCount": 7721, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/ca235ce0decdb4f80024a429a20ae4437ceae09e", "doi": "10.1109/CVPR.2019.00482", "arxiv": "" }, { "title": "Structure-from-Motion Revisited", "authors": "Johannes L. Schönberger, Jan-Michael Frahm", "venue": "CVPR", "year": 2016, "topic": "3D / Geometry / Driving", "modality": "3D / depth / optical flow / driving sensors", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "三维理解、几何重建、光流、深度或自动驾驶感知", "method": "几何约束、点云网络、连续场表示或多传感器数据集", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "This paper revisits incremental structure-from-motion and identifies practical bottlenecks in robustness, scalability, initialization, triangulation, image registration, and bundle adjustment. It presents a complete SfM system that improves these steps and serves as the basis of COLMAP.", "motivation": "Incremental SfM is accurate but fragile at scale: image registration, initialization, triangulation, and accumulated reconstruction errors can make large unordered photo collections hard to reconstruct reliably.", "implementation": "The work redesigns the SfM pipeline with more robust feature matching and verification, improved image registration and triangulation, and scalable bundle-adjustment choices in a complete reconstruction system.", "application": "Useful for large-scale 3D reconstruction from unordered image collections, camera pose estimation, photogrammetry, mapping, and downstream neural rendering or scene understanding pipelines.", "problem": "Incremental SfM is accurate but fragile at scale: image registration, initialization, triangulation, and accumulated reconstruction errors can make large unordered photo collections hard to reconstruct reliably.", "contribution": "The work redesigns the SfM pipeline with more robust feature matching and verification, improved image registration and triangulation, and scalable bundle-adjustment choices in a complete reconstruction system.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Structure-from-Motion Revisited 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 3D / Geometry / Driving 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 7,314；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何把真实传感噪声、动态场景和跨域泛化纳入可靠评测。", "directionNote": "Useful for large-scale 3D reconstruction from unordered image collections, camera pose estimation, photogrammetry, mapping, and downstream neural rendering or scene understanding pipelines.", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "3D / Geometry / Driving" ], "code": { "available": false, "url": "" }, "url": "https://www.semanticscholar.org/paper/e7d0c37f4f3589a3b787f39e8307704da5ed8d6c", "updated": "2026-05-20", "topCitedRank": 30, "citationCount": 7314, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/e7d0c37f4f3589a3b787f39e8307704da5ed8d6c", "doi": "10.1109/CVPR.2016.445", "arxiv": "" }, { "title": "Realtime Multi-person 2D Pose Estimation Using Part Affinity Fields", "authors": "Zhe Cao, T. Simon, S. Wei et al.", "venue": "CVPR", "year": 2016, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "We present an approach to efficiently detect the 2D pose of multiple people in an image. The approach uses a nonparametric representation, which we refer to as Part Affinity Fields (PAFs), to learn to associate body parts with individuals in the image. The architecture encodes global context, allowing a greedy bottom-up parsing step that maintains high accuracy while achieving realtime performance, irrespective of the number of people in the image. The architecture is designed to jointly learn part locations and their association via two branches of the same sequential prediction process. Our method placed first in the inaugural COCO 2016 keypoints challenge, and significantly exceeds the previous state-of-the-art result on the MPII Multi-Person benchmark, both in performance and efficiency.", "motivation": "Our method placed first in the inaugural COCO 2016 keypoints challenge, and significantly exceeds the previous state-of-the-art result on the MPII Multi-Person benchmark, both in performance and efficiency.", "implementation": "The architecture encodes global context, allowing a greedy bottom-up parsing step that maintains high accuracy while achieving realtime performance, irrespective of the number of people in the image.", "application": "The architecture is designed to jointly learn part locations and their association via two branches of the same sequential prediction process.", "problem": "Our method placed first in the inaugural COCO 2016 keypoints challenge, and significantly exceeds the previous state-of-the-art result on the MPII Multi-Person benchmark, both in performance and efficiency.", "contribution": "The architecture encodes global context, allowing a greedy bottom-up parsing step that maintains high accuracy while achieving realtime performance, irrespective of the number of people in the image.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Realtime Multi-person 2D Pose Estimation Using Part Affinity Fields 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 7,256；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Datasets / Benchmarks" ], "code": { "available": false, "url": "" }, "url": "https://arxiv.org/pdf/1611.08050", "updated": "2026-05-20", "topCitedRank": 31, "citationCount": 7256, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/dfad8f616bd2a05c8cae5f61060f743f966ece85", "doi": "10.1109/CVPR.2017.143", "arxiv": "1611.08050" }, { "title": "Analyzing and Improving the Image Quality of StyleGAN", "authors": "Tero Karras, S. Laine, M. Aittala et al.", "venue": "CVPR", "year": 2019, "topic": "Generative Vision", "modality": "Image generation / editing", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "图像生成、编辑、翻译或高保真合成", "method": "GAN、diffusion、transformer 或条件生成模型", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "The style-based GAN architecture (StyleGAN) yields state-of-the-art results in data-driven unconditional generative image modeling. We expose and analyze several of its characteristic artifacts, and propose changes in both model architecture and training methods to address them. In particular, we redesign the generator normalization, revisit progressive growing, and regularize the generator to encourage good conditioning in the mapping from latent codes to images. In addition to improving image quality, this path length regularizer yields the additional benefit that the generator becomes significantly easier to invert. This makes it possible to reliably attribute a generated image to a particular network. We furthermore visualize how well the generator utilizes its output resolution, and identify a capacity problem, motivating us to train larger models for additional quality improvements. Overall, our improved model redefines the state of the art in unconditional image modeling, both in terms of existing distribution quality metrics as well as perceived image quality.", "motivation": "We furthermore visualize how well the generator utilizes its output resolution, and identify a capacity problem, motivating us to train larger models for additional quality improvements.", "implementation": "The style-based GAN architecture (StyleGAN) yields state-of-the-art results in data-driven unconditional generative image modeling.", "application": "In addition to improving image quality, this path length regularizer yields the additional benefit that the generator becomes significantly easier to invert.", "problem": "We furthermore visualize how well the generator utilizes its output resolution, and identify a capacity problem, motivating us to train larger models for additional quality improvements.", "contribution": "The style-based GAN architecture (StyleGAN) yields state-of-the-art results in data-driven unconditional generative image modeling.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Analyzing and Improving the Image Quality of StyleGAN 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Generative Vision 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 6,938；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何平衡生成质量、可控性、版权风险和真实任务可靠性。", "directionNote": "该方向的高引用工作集中体现了 Generative Vision 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Generative Vision" ], "code": { "available": false, "url": "" }, "url": "https://arxiv.org/pdf/1912.04958", "updated": "2026-05-20", "topCitedRank": 32, "citationCount": 6938, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/14fdc18d9c164e5b0d6d946b3238c04e81921358", "doi": "10.1109/cvpr42600.2020.00813", "arxiv": "1912.04958" }, { "title": "Adding Conditional Control to Text-to-Image Diffusion Models", "authors": "Lvmin Zhang, Anyi Rao, Maneesh Agrawala", "venue": "ICCV", "year": 2023, "topic": "Vision-Language / Foundation Models", "modality": "Image-text / foundation model", "status": "CV 顶会高引 Top 100", "priority": "High", "task": "视觉-语言理解、开放词表识别或基础模型适配", "method": "大规模预训练、多模态对齐或提示式视觉建模", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "We present ControlNet, a neural network architecture to add spatial conditioning controls to large, pretrained text-to-image diffusion models. ControlNet locks the production-ready large diffusion models, and reuses their deep and robust encoding layers pretrained with billions of images as a strong backbone to learn a diverse set of conditional controls. The neural architecture is connected with \"zero convolutions\" (zero-initialized convolution layers) that progressively grow the parameters from zero and ensure that no harmful noise could affect the finetuning. We test various conditioning controls, e.g., edges, depth, segmentation, human pose, etc., with Stable Diffusion, using single or multiple conditions, with or without prompts. We show that the training of ControlNets is robust with small (<50k) and large (>1m) datasets. Extensive results show that ControlNet may facilitate wider applications to control image diffusion models.", "motivation": "We present ControlNet, a neural network architecture to add spatial conditioning controls to large, pretrained text-to-image diffusion models.", "implementation": "The neural architecture is connected with \"zero convolutions\" (zero-initialized convolution layers) that progressively grow the parameters from zero and ensure that no harmful noise could affect the finetuning.", "application": "We test various conditioning controls, e.g., edges, depth, segmentation, human pose, etc., with Stable Diffusion, using single or multiple conditions, with or without prompts.", "problem": "We present ControlNet, a neural network architecture to add spatial conditioning controls to large, pretrained text-to-image diffusion models.", "contribution": "The neural architecture is connected with \"zero convolutions\" (zero-initialized convolution layers) that progressively grow the parameters from zero and ensure that no harmful noise could affect the finetuning.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Adding Conditional Control to Text-to-Image Diffusion Models 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Vision-Language / Foundation Models 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 6,857；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何降低标注和算力成本，同时保持可控性、可解释性和领域泛化。", "directionNote": "该方向的高引用工作集中体现了 Vision-Language / Foundation Models 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "ICCV", "Vision-Language / Foundation Models", "award-winning", "Best Paper" ], "code": { "available": false, "url": "" }, "url": "https://arxiv.org/pdf/2302.05543", "updated": "2026-05-20", "topCitedRank": 33, "citationCount": 6857, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/efbe97d20c4ffe356e8826c01dc550bacc405add", "doi": "10.1109/ICCV51070.2023.00355", "arxiv": "2302.05543", "awards": [ { "conference": "ICCV", "year": 2023, "award": "Best Paper", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ] }, { "title": "EfficientDet: Scalable and Efficient Object Detection", "authors": "Mingxing Tan, Ruoming Pang, Quoc V. Le", "venue": "CVPR", "year": 2019, "topic": "Object Detection", "modality": "RGB / multi-scale visual features", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "目标检测、开放集检测或检测损失设计", "method": "检测框架、特征金字塔、transformer detector 或检测损失", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "Model efficiency has become increasingly important in computer vision. In this paper, we systematically study neural network architecture design choices for object detection and propose several key optimizations to improve efficiency. First, we propose a weighted bi-directional feature pyramid network (BiFPN), which allows easy and fast multi-scale feature fusion; Second, we propose a compound scaling method that uniformly scales the resolution, depth, and width for all backbone, feature network, and box/class prediction networks at the same time. Based on these optimizations and EfficientNet backbones, we have developed a new family of object detectors, called EfficientDet, which consistently achieve much better efficiency than prior art across a wide spectrum of resource constraints. In particular, with single-model and single-scale, our EfficientDetD7 achieves state-of-the-art 52.2 AP on COCO test-dev with 52M parameters and 325B FLOPs, being 4x – 9x smaller and using 13x – 42x fewer FLOPs than previous detector.", "motivation": "Model efficiency has become increasingly important in computer vision.", "implementation": "In this paper, we systematically study neural network architecture design choices for object detection and propose several key optimizations to improve efficiency.", "application": "In particular, with single-model and single-scale, our EfficientDetD7 achieves state-of-the-art 52.2 AP on COCO test-dev with 52M parameters and 325B FLOPs, being 4x – 9x smaller and using 13x – 42x fewer FLOPs than previous detector.", "problem": "Model efficiency has become increasingly important in computer vision.", "contribution": "In this paper, we systematically study neural network architecture design choices for object detection and propose several key optimizations to improve efficiency.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "EfficientDet: Scalable and Efficient Object Detection 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Object Detection 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 6,836；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何在开放世界、长尾类别和实时部署中保持稳定检测性能。", "directionNote": "该方向的高引用工作集中体现了 Object Detection 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Object Detection" ], "code": { "available": false, "url": "" }, "url": "https://arxiv.org/pdf/1911.09070", "updated": "2026-05-20", "topCitedRank": 34, "citationCount": 6836, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/41c67d04be2d1632c0d3b0880c21c9fe797cdab8", "doi": "10.1109/cvpr42600.2020.01079", "arxiv": "1911.09070" }, { "title": "Deformable Convolutional Networks", "authors": "Jifeng Dai, Haozhi Qi, Yuwen Xiong et al.", "venue": "ICCV", "year": 2017, "topic": "Object Detection", "modality": "RGB / multi-scale visual features", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "目标检测、开放集检测或检测损失设计", "method": "检测框架、特征金字塔、transformer detector 或检测损失", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "Convolutional neural networks (CNNs) are inherently limited to model geometric transformations due to the fixed geometric structures in their building modules. In this work, we introduce two new modules to enhance the transformation modeling capability of CNNs, namely, deformable convolution and deformable RoI pooling. Both are based on the idea of augmenting the spatial sampling locations in the modules with additional offsets and learning the offsets from the target tasks, without additional supervision. The new modules can readily replace their plain counterparts in existing CNNs and can be easily trained end-to-end by standard back-propagation, giving rise to deformable convolutional networks. Extensive experiments validate the performance of our approach. For the first time, we show that learning dense spatial transformation in deep CNNs is effective for sophisticated vision tasks such as object detection and semantic segmentation. The code is released at https://github.com/msracver/Deformable-ConvNets.", "motivation": "Convolutional neural networks (CNNs) are inherently limited to model geometric transformations due to the fixed geometric structures in their building modules.", "implementation": "In this work, we introduce two new modules to enhance the transformation modeling capability of CNNs, namely, deformable convolution and deformable RoI pooling.", "application": "For the first time, we show that learning dense spatial transformation in deep CNNs is effective for sophisticated vision tasks such as object detection and semantic segmentation.", "problem": "Convolutional neural networks (CNNs) are inherently limited to model geometric transformations due to the fixed geometric structures in their building modules.", "contribution": "In this work, we introduce two new modules to enhance the transformation modeling capability of CNNs, namely, deformable convolution and deformable RoI pooling.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Deformable Convolutional Networks 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Object Detection 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 6,493；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何在开放世界、长尾类别和实时部署中保持稳定检测性能。", "directionNote": "该方向的高引用工作集中体现了 Object Detection 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "ICCV", "Object Detection" ], "code": { "available": true, "url": "https://github.com/msracver/Deformable-ConvNets" }, "url": "https://arxiv.org/pdf/1703.06211", "updated": "2026-05-20", "topCitedRank": 35, "citationCount": 6493, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/4a73a1840945e87583d89ca0216a2c449d50a4a3", "doi": "10.1109/ICCV.2017.89", "arxiv": "1703.06211" }, { "title": "ShuffleNet V2: Practical Guidelines for Efficient CNN Architecture Design", "authors": "Ningning Ma, Xiangyu Zhang, Haitao Zheng et al.", "venue": "ECCV", "year": 2018, "topic": "Video / Human Understanding", "modality": "Video / human-centric data", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "视频理解、动作识别、人体姿态或轨迹预测", "method": "时序建模、人体结构先验或视频 transformer", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "Currently, the neural network architecture design is mostly guided by the indirect metric of computation complexity, i.e., FLOPs. However, the direct metric, e.g., speed, also depends on the other factors such as memory access cost and platform characterics. Thus, this work proposes to evaluate the direct metric on the target platform, beyond only considering FLOPs. Based on a series of controlled experiments, this work derives several practical guidelines for efficient network design. Accordingly, a new architecture is presented, called ShuffleNet V2. Comprehensive ablation experiments verify that our model is the state-of-the-art in terms of speed and accuracy tradeoff.", "motivation": "However, the direct metric, e.g., speed, also depends on the other factors such as memory access cost and platform characterics.", "implementation": "Currently, the neural network architecture design is mostly guided by the indirect metric of computation complexity, i.e., FLOPs.", "application": "Thus, this work proposes to evaluate the direct metric on the target platform, beyond only considering FLOPs.", "problem": "However, the direct metric, e.g., speed, also depends on the other factors such as memory access cost and platform characterics.", "contribution": "Currently, the neural network architecture design is mostly guided by the indirect metric of computation complexity, i.e., FLOPs.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "ShuffleNet V2: Practical Guidelines for Efficient CNN Architecture Design 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Video / Human Understanding 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 6,284；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何在长时序、遮挡和复杂交互下保持鲁棒理解。", "directionNote": "该方向的高引用工作集中体现了 Video / Human Understanding 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "ECCV", "Video / Human Understanding" ], "code": { "available": false, "url": "" }, "url": "https://www.semanticscholar.org/paper/c02b909a514af6b9255315e2d50112845ca5ed0e", "updated": "2026-05-20", "topCitedRank": 36, "citationCount": 6284, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/c02b909a514af6b9255315e2d50112845ca5ed0e", "doi": "10.1007/978-3-030-01264-9_8", "arxiv": "1807.11164" }, { "title": "FCOS: Fully Convolutional One-Stage Object Detection", "authors": "Zhi Tian, Chunhua Shen, Hao Chen et al.", "venue": "ICCV", "year": 2019, "topic": "Object Detection", "modality": "RGB / multi-scale visual features", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "目标检测、开放集检测或检测损失设计", "method": "检测框架、特征金字塔、transformer detector 或检测损失", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "We propose a fully convolutional one-stage object detector (FCOS) to solve object detection in a per-pixel prediction fashion, analogue to semantic segmentation. Almost all state-of-the-art object detectors such as RetinaNet, SSD, YOLOv3, and Faster R-CNN rely on pre-defined anchor boxes. In contrast, our proposed detector FCOS is anchor box free, as well as proposal free. By eliminating the pre-defined set of anchor boxes, FCOS completely avoids the complicated computation related to anchor boxes such as calculating overlapping during training. More importantly, we also avoid all hyper-parameters related to anchor boxes, which are often very sensitive to the final detection performance. With the only post-processing non-maximum suppression (NMS), FCOS with ResNeXt-64x4d-101 achieves 44.7% in AP with single-model and single-scale testing, surpassing previous one-stage detectors with the advantage of being much simpler. For the first time, we demonstrate a much simpler and flexible detection framework achieving improved detection accuracy. We hope that the proposed FCOS framework can serve as a simple and strong alternative for many other instance-level tasks. Code is available at: https://tinyurl.com/FCOSv1", "motivation": "We propose a fully convolutional one-stage object detector (FCOS) to solve object detection in a per-pixel prediction fashion, analogue to semantic segmentation.", "implementation": "With the only post-processing non-maximum suppression (NMS), FCOS with ResNeXt-64x4d-101 achieves 44.7% in AP with single-model and single-scale testing, surpassing previous one-stage detectors with the advantage of being much simpler.", "application": "More importantly, we also avoid all hyper-parameters related to anchor boxes, which are often very sensitive to the final detection performance.", "problem": "We propose a fully convolutional one-stage object detector (FCOS) to solve object detection in a per-pixel prediction fashion, analogue to semantic segmentation.", "contribution": "With the only post-processing non-maximum suppression (NMS), FCOS with ResNeXt-64x4d-101 achieves 44.7% in AP with single-model and single-scale testing, surpassing previous one-stage detectors with the advantage of being much simpler.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "FCOS: Fully Convolutional One-Stage Object Detection 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Object Detection 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 6,087；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何在开放世界、长尾类别和实时部署中保持稳定检测性能。", "directionNote": "该方向的高引用工作集中体现了 Object Detection 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "ICCV", "Object Detection" ], "code": { "available": false, "url": "" }, "url": "https://arxiv.org/pdf/1904.01355", "updated": "2026-05-20", "topCitedRank": 37, "citationCount": 6087, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/e2751a898867ce6687e08a5cc7bdb562e999b841", "doi": "10.1109/ICCV.2019.00972", "arxiv": "1904.01355" }, { "title": "Dual Attention Network for Scene Segmentation", "authors": "J. Fu, J. Liu, Haijie Tian et al.", "venue": "CVPR", "year": 2018, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "In this paper, we address the scene segmentation task by capturing rich contextual dependencies based on the self-attention mechanism. Unlike previous works that capture contexts by multi-scale features fusion, we propose a Dual Attention Networks (DANet) to adaptively integrate local features with their global dependencies. Specifically, we append two types of attention modules on top of traditional dilated FCN, which model the semantic interdependencies in spatial and channel dimensions respectively. The position attention module selectively aggregates the features at each position by a weighted sum of the features at all positions. Similar features would be related to each other regardless of their distances. Meanwhile, the channel attention module selectively emphasizes interdependent channel maps by integrating associated features among all channel maps. We sum the outputs of the two attention modules to further improve feature representation which contributes to more precise segmentation results. We achieve new state-of-the-art segmentation performance on three challenging scene segmentation datasets, i.e., Cityscapes, PASCAL Context and COCO Stuff dataset. In particular, a Mean IoU score of 81.5% on Cityscapes test set is achieved without using coarse data.", "motivation": "We achieve new state-of-the-art segmentation performance on three challenging scene segmentation datasets, i.e., Cityscapes, PASCAL Context and COCO Stuff dataset.", "implementation": "Unlike previous works that capture contexts by multi-scale features fusion, we propose a Dual Attention Networks (DANet) to adaptively integrate local features with their global dependencies.", "application": "In this paper, we address the scene segmentation task by capturing rich contextual dependencies based on the self-attention mechanism.", "problem": "We achieve new state-of-the-art segmentation performance on three challenging scene segmentation datasets, i.e., Cityscapes, PASCAL Context and COCO Stuff dataset.", "contribution": "Unlike previous works that capture contexts by multi-scale features fusion, we propose a Dual Attention Networks (DANet) to adaptively integrate local features with their global dependencies.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Dual Attention Network for Scene Segmentation 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 6,015；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Datasets / Benchmarks" ], "code": { "available": false, "url": "" }, "url": "https://arxiv.org/pdf/1809.02983", "updated": "2026-05-20", "topCitedRank": 38, "citationCount": 6015, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/ad655c25e052fa4eeed53421344aca6f239c4c9d", "doi": "10.1109/CVPR.2019.00326", "arxiv": "1809.02983" }, { "title": "Real-Time Single Image and Video Super-Resolution Using an Efficient Sub-Pixel Convolutional Neural Network", "authors": "Wenzhe Shi, Jose Caballero, Ferenc Huszár et al.", "venue": "CVPR", "year": 2016, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "Recently, several models based on deep neural networks have achieved great success in terms of both reconstruction accuracy and computational performance for single image super-resolution. In these methods, the low resolution (LR) input image is upscaled to the high resolution (HR) space using a single filter, commonly bicubic interpolation, before reconstruction. This means that the super-resolution (SR) operation is performed in HR space. We demonstrate that this is sub-optimal and adds computational complexity. In this paper, we present the first convolutional neural network (CNN) capable of real-time SR of 1080p videos on a single K2 GPU. To achieve this, we propose a novel CNN architecture where the feature maps are extracted in the LR space. In addition, we introduce an efficient sub-pixel convolution layer which learns an array of upscaling filters to upscale the final LR feature maps into the HR output. By doing so, we effectively replace the handcrafted bicubic filter in the SR pipeline with more complex upscaling filters specifically trained for each feature map, whilst also reducing the computational complexity of the overall SR operation. We evaluate the proposed approach using images and videos from publicly available datasets and show that it performs significantly better (+0.15dB on Images and +0.39dB on Videos) and is an order of magnitude faster than previous CNN-based methods.", "motivation": "Recently, several models based on deep neural networks have achieved great success in terms of both reconstruction accuracy and computational performance for single image super-resolution.", "implementation": "In this paper, we present the first convolutional neural network (CNN) capable of real-time SR of 1080p videos on a single K2 GPU.", "application": "In these methods, the low resolution (LR) input image is upscaled to the high resolution (HR) space using a single filter, commonly bicubic interpolation, before reconstruction.", "problem": "Recently, several models based on deep neural networks have achieved great success in terms of both reconstruction accuracy and computational performance for single image super-resolution.", "contribution": "In this paper, we present the first convolutional neural network (CNN) capable of real-time SR of 1080p videos on a single K2 GPU.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Real-Time Single Image and Video Super-Resolution Using an Efficient Sub-Pixel Convolutional Neural Network 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 5,919；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Datasets / Benchmarks" ], "code": { "available": false, "url": "" }, "url": "http://arxiv.org/pdf/1609.05158", "updated": "2026-05-20", "topCitedRank": 39, "citationCount": 5919, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/03a5b2aac53443e6078f0f63b35d4f95d6d54c5d", "doi": "10.1109/CVPR.2016.207", "arxiv": "1609.05158" }, { "title": "ECA-Net: Efficient Channel Attention for Deep Convolutional Neural Networks", "authors": "Qilong Wang, Banggu Wu, Peng Fei Zhu et al.", "venue": "CVPR", "year": 2019, "topic": "Object Detection", "modality": "RGB / multi-scale visual features", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "目标检测、开放集检测或检测损失设计", "method": "检测框架、特征金字塔、transformer detector 或检测损失", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "Recently, channel attention mechanism has demonstrated to offer great potential in improving the performance of deep convolutional neural networks (CNNs). However, most existing methods dedicate to developing more sophisticated attention modules for achieving better performance, which inevitably increase model complexity. To overcome the paradox of performance and complexity trade-off, this paper proposes an Efficient Channel Attention (ECA) module, which only involves a handful of parameters while bringing clear performance gain. By dissecting the channel attention module in SENet, we empirically show avoiding dimensionality reduction is important for learning channel attention, and appropriate cross-channel interaction can preserve performance while significantly decreasing model complexity. Therefore, we propose a local cross-channel interaction strategy without dimensionality reduction, which can be efficiently implemented via 1D convolution. Furthermore, we develop a method to adaptively select kernel size of 1D convolution, determining coverage of local cross-channel interaction. The proposed ECA module is both efficient and effective, e.g., the parameters and computations of our modules against backbone of ResNet50 are 80 vs. 24.37M and 4.7e-4 GFlops vs. 3.86 GFlops, respectively, and the performance boost is more than 2% in terms of Top-1 accuracy. We extensively evaluate our ECA module on image classification, object detection and instance segmentation with backbones of ResNets and MobileNetV2. The experimental results show our module is more efficient while performing favorably against its counterparts.", "motivation": "However, most existing methods dedicate to developing more sophisticated attention modules for achieving better performance, which inevitably increase model complexity.", "implementation": "To overcome the paradox of performance and complexity trade-off, this paper proposes an Efficient Channel Attention (ECA) module, which only involves a handful of parameters while bringing clear performance gain.", "application": "We extensively evaluate our ECA module on image classification, object detection and instance segmentation with backbones of ResNets and MobileNetV2.", "problem": "However, most existing methods dedicate to developing more sophisticated attention modules for achieving better performance, which inevitably increase model complexity.", "contribution": "To overcome the paradox of performance and complexity trade-off, this paper proposes an Efficient Channel Attention (ECA) module, which only involves a handful of parameters while bringing clear performance gain.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "ECA-Net: Efficient Channel Attention for Deep Convolutional Neural Networks 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Object Detection 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 5,883；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何在开放世界、长尾类别和实时部署中保持稳定检测性能。", "directionNote": "该方向的高引用工作集中体现了 Object Detection 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Object Detection" ], "code": { "available": false, "url": "" }, "url": "https://arxiv.org/pdf/1910.03151", "updated": "2026-05-20", "topCitedRank": 40, "citationCount": 5883, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/8cb34cbdcf65c23ef98430441b14a648c4e8d992", "doi": "10.1109/CVPR42600.2020.01155", "arxiv": "1910.03151" }, { "title": "CutMix: Regularization Strategy to Train Strong Classifiers With Localizable Features", "authors": "Sangdoo Yun, Dongyoon Han, Seong Joon Oh et al.", "venue": "ICCV", "year": 2019, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "Regional dropout strategies have been proposed to enhance performance of convolutional neural network classifiers. They have proved to be effective for guiding the model to attend on less discriminative parts of objects (e.g. leg as opposed to head of a person), thereby letting the network generalize better and have better object localization capabilities. On the other hand, current methods for regional dropout removes informative pixels on training images by overlaying a patch of either black pixels or random noise. Such removal is not desirable because it suffers from information loss causing inefficiency in training. We therefore propose the CutMix augmentation strategy: patches are cut and pasted among training images where the ground truth labels are also mixed proportionally to the area of the patches. By making efficient use of training pixels and retaining the regularization effect of regional dropout, CutMix consistently outperforms state-of-the-art augmentation strategies on CIFAR and ImageNet classification tasks, as well as on ImageNet weakly-supervised localization task. Moreover, unlike previous augmentation methods, our CutMix-trained ImageNet classifier, when used as a pretrained model, results in consistent performance gain in Pascal detection and MS-COCO image captioning benchmarks. We also show that CutMix can improve the model robustness against input corruptions and its out-of distribution detection performance.", "motivation": "Regional dropout strategies have been proposed to enhance performance of convolutional neural network classifiers.", "implementation": "They have proved to be effective for guiding the model to attend on less discriminative parts of objects (e.g.", "application": "By making efficient use of training pixels and retaining the regularization effect of regional dropout, CutMix consistently outperforms state-of-the-art augmentation strategies on CIFAR and ImageNet classification tasks, as well as on ImageNet weakly-supervised localization task.", "problem": "Regional dropout strategies have been proposed to enhance performance of convolutional neural network classifiers.", "contribution": "They have proved to be effective for guiding the model to attend on less discriminative parts of objects (e.g.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "CutMix: Regularization Strategy to Train Strong Classifiers With Localizable Features 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 5,878；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "ICCV", "Datasets / Benchmarks" ], "code": { "available": false, "url": "" }, "url": "https://arxiv.org/pdf/1905.04899", "updated": "2026-05-20", "topCitedRank": 41, "citationCount": 5878, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/ed17929e66da7f8fbc3666bf5eb613d302ddde0c", "doi": "10.1109/ICCV.2019.00612", "arxiv": "1905.04899" }, { "title": "Scalable Diffusion Models with Transformers", "authors": "William S. Peebles, Saining Xie", "venue": "ICCV", "year": 2022, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "We explore a new class of diffusion models based on the transformer architecture. We train latent diffusion models of images, replacing the commonly-used U-Net backbone with a transformer that operates on latent patches. We analyze the scalability of our Diffusion Transformers (DiTs) through the lens of forward pass complexity as measured by Gflops. We find that DiTs with higher Gflops—through increased transformer depth/width or increased number of input tokens—consistently have lower FID. In addition to possessing good scalability properties, our largest DiT-XL/2 models outperform all prior diffusion models on the class-conditional ImageNet 512×512 and 256×256 benchmarks, achieving a state-of-the-art FID of 2.27 on the latter.", "motivation": "We explore a new class of diffusion models based on the transformer architecture.", "implementation": "We analyze the scalability of our Diffusion Transformers (DiTs) through the lens of forward pass complexity as measured by Gflops.", "application": "In addition to possessing good scalability properties, our largest DiT-XL/2 models outperform all prior diffusion models on the class-conditional ImageNet 512×512 and 256×256 benchmarks, achieving a state-of-the-art FID of 2.27 on the latter.", "problem": "We explore a new class of diffusion models based on the transformer architecture.", "contribution": "We analyze the scalability of our Diffusion Transformers (DiTs) through the lens of forward pass complexity as measured by Gflops.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Scalable Diffusion Models with Transformers 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 5,718；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "ICCV", "Datasets / Benchmarks" ], "code": { "available": false, "url": "" }, "url": "https://arxiv.org/pdf/2212.09748", "updated": "2026-05-20", "topCitedRank": 42, "citationCount": 5718, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/736973165f98105fec3729b7db414ae4d80fcbeb", "doi": "10.1109/ICCV51070.2023.00387", "arxiv": "2212.09748" }, { "title": "Context Encoders: Feature Learning by Inpainting", "authors": "Deepak Pathak, Philipp Krähenbühl, Jeff Donahue et al.", "venue": "CVPR", "year": 2016, "topic": "Generative Vision", "modality": "Image generation / editing", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "图像生成、编辑、翻译或高保真合成", "method": "GAN、diffusion、transformer 或条件生成模型", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "We present an unsupervised visual feature learning algorithm driven by context-based pixel prediction. By analogy with auto-encoders, we propose Context Encoders - a convolutional neural network trained to generate the contents of an arbitrary image region conditioned on its surroundings. In order to succeed at this task, context encoders need to both understand the content of the entire image, as well as produce a plausible hypothesis for the missing part(s). When training context encoders, we have experimented with both a standard pixel-wise reconstruction loss, as well as a reconstruction plus an adversarial loss. The latter produces much sharper results because it can better handle multiple modes in the output. We found that a context encoder learns a representation that captures not just appearance but also the semantics of visual structures. We quantitatively demonstrate the effectiveness of our learned features for CNN pre-training on classification, detection, and segmentation tasks. Furthermore, context encoders can be used for semantic inpainting tasks, either stand-alone or as initialization for non-parametric methods.", "motivation": "In order to succeed at this task, context encoders need to both understand the content of the entire image, as well as produce a plausible hypothesis for the missing part(s).", "implementation": "We present an unsupervised visual feature learning algorithm driven by context-based pixel prediction.", "application": "When training context encoders, we have experimented with both a standard pixel-wise reconstruction loss, as well as a reconstruction plus an adversarial loss.", "problem": "In order to succeed at this task, context encoders need to both understand the content of the entire image, as well as produce a plausible hypothesis for the missing part(s).", "contribution": "We present an unsupervised visual feature learning algorithm driven by context-based pixel prediction.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Context Encoders: Feature Learning by Inpainting 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Generative Vision 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 5,717；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何平衡生成质量、可控性、版权风险和真实任务可靠性。", "directionNote": "该方向的高引用工作集中体现了 Generative Vision 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Generative Vision" ], "code": { "available": false, "url": "" }, "url": "http://arxiv.org/pdf/1604.07379", "updated": "2026-05-20", "topCitedRank": 43, "citationCount": 5717, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/7d0effebfa4bed19b6ba41f3af5b7e5b6890de87", "doi": "10.1109/CVPR.2016.278", "arxiv": "1604.07379" }, { "title": "Image Style Transfer Using Convolutional Neural Networks", "authors": "Leon A. Gatys, Alexander S. Ecker, M. Bethge", "venue": "CVPR", "year": 2016, "topic": "Generative Vision", "modality": "Image generation / editing", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "图像生成、编辑、翻译或高保真合成", "method": "GAN、diffusion、transformer 或条件生成模型", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "This paper shows that convolutional neural networks can separate image content and style representations. It synthesizes an image by optimizing pixels to match high-level content features from one image and style statistics from another.", "motivation": "Artistic stylization requires preserving the semantic content of one image while transferring the visual texture, color, and brush-stroke statistics of another, which traditional filters do not model explicitly.", "implementation": "The method uses CNN activations for content representation and Gram-matrix correlations of feature maps for style representation, then optimizes a generated image to jointly match both losses.", "application": "Useful for neural artistic style transfer, image editing, texture synthesis, creative visual effects, and studying feature representations learned by classification CNNs.", "problem": "Artistic stylization requires preserving the semantic content of one image while transferring the visual texture, color, and brush-stroke statistics of another, which traditional filters do not model explicitly.", "contribution": "The method uses CNN activations for content representation and Gram-matrix correlations of feature maps for style representation, then optimizes a generated image to jointly match both losses.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Image Style Transfer Using Convolutional Neural Networks 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Generative Vision 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 5,711；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何平衡生成质量、可控性、版权风险和真实任务可靠性。", "directionNote": "Useful for neural artistic style transfer, image editing, texture synthesis, creative visual effects, and studying feature representations learned by classification CNNs.", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Generative Vision" ], "code": { "available": false, "url": "" }, "url": "https://www.semanticscholar.org/paper/7568d13a82f7afa4be79f09c295940e48ec6db89", "updated": "2026-05-20", "topCitedRank": 44, "citationCount": 5711, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/7568d13a82f7afa4be79f09c295940e48ec6db89", "doi": "10.1109/CVPR.2016.265", "arxiv": "" }, { "title": "Unpaired Image-to-Image Translation Using Cycle-Consistent Adversarial Networks", "authors": "Jun-Yan Zhu, Taesung Park, Phillip Isola et al.", "venue": "ICCV", "year": 2017, "topic": "Generative Vision", "modality": "Image generation / editing", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "图像生成、编辑、翻译或高保真合成", "method": "GAN、diffusion、transformer 或条件生成模型", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "Image-to-image translation is a class of vision and graphics problems where the goal is to learn the mapping between an input image and an output image using a training set of aligned image pairs. However, for many tasks, paired training data will not be available. We present an approach for learning to translate an image from a source domain X to a target domain Y in the absence of paired examples. Our goal is to learn a mapping G : X → Y such that the distribution of images from G(X) is indistinguishable from the distribution Y using an adversarial loss. Because this mapping is highly under-constrained, we couple it with an inverse mapping F : Y → X and introduce a cycle consistency loss to push F(G(X)) ≈ X (and vice versa). Qualitative results are presented on several tasks where paired training data does not exist, including collection style transfer, object transfiguration, season transfer, photo enhancement, etc. Quantitative comparisons against several prior methods demonstrate the superiority of our approach.", "motivation": "However, for many tasks, paired training data will not be available.", "implementation": "Our goal is to learn a mapping G : X → Y such that the distribution of images from G(X) is indistinguishable from the distribution Y using an adversarial loss.", "application": "Because this mapping is highly under-constrained, we couple it with an inverse mapping F : Y → X and introduce a cycle consistency loss to push F(G(X)) ≈ X (and vice versa).", "problem": "However, for many tasks, paired training data will not be available.", "contribution": "Our goal is to learn a mapping G : X → Y such that the distribution of images from G(X) is indistinguishable from the distribution Y using an adversarial loss.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Unpaired Image-to-Image Translation Using Cycle-Consistent Adversarial Networks 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Generative Vision 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 5,644；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何平衡生成质量、可控性、版权风险和真实任务可靠性。", "directionNote": "该方向的高引用工作集中体现了 Generative Vision 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "ICCV", "Generative Vision" ], "code": { "available": false, "url": "" }, "url": "https://repositorio.unal.edu.co/bitstream/unal/82529/2/98562187.2022.pdf", "updated": "2026-05-20", "topCitedRank": 45, "citationCount": 5644, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/c43d954cf8133e6254499f3d68e45218067e4941", "doi": "10.1109/ICCV.2017.244", "arxiv": "1703.10593" }, { "title": "Generalized Intersection Over Union: A Metric and a Loss for Bounding Box Regression", "authors": "S. H. Rezatofighi, Nathan Tsoi, JunYoung Gwak et al.", "venue": "CVPR", "year": 2019, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "Intersection over Union (IoU) is the most popular evaluation metric used in the object detection benchmarks. However, there is a gap between optimizing the commonly used distance losses for regressing the parameters of a bounding box and maximizing this metric value. The optimal objective for a metric is the metric itself. In the case of axis-aligned 2D bounding boxes, it can be shown that IoU can be directly used as a regression loss. However, IoU has a plateau making it infeasible to optimize in the case of non-overlapping bounding boxes. In this paper, we address the this weakness by introducing a generalized version of IoU as both a new loss and a new metric. By incorporating this generalized IoU ( GIoU) as a loss into the state-of-the art object detection frameworks, we show a consistent improvement on their performance using both the standard, IoU based, and new, GIoU based, performance measures on popular object detection benchmarks such as PASCAL VOC and MS COCO.", "motivation": "However, there is a gap between optimizing the commonly used distance losses for regressing the parameters of a bounding box and maximizing this metric value.", "implementation": "In the case of axis-aligned 2D bounding boxes, it can be shown that IoU can be directly used as a regression loss.", "application": "Intersection over Union (IoU) is the most popular evaluation metric used in the object detection benchmarks.", "problem": "However, there is a gap between optimizing the commonly used distance losses for regressing the parameters of a bounding box and maximizing this metric value.", "contribution": "In the case of axis-aligned 2D bounding boxes, it can be shown that IoU can be directly used as a regression loss.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Generalized Intersection Over Union: A Metric and a Loss for Bounding Box Regression 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 5,558；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Datasets / Benchmarks" ], "code": { "available": false, "url": "" }, "url": "https://arxiv.org/pdf/1902.09630", "updated": "2026-05-20", "topCitedRank": 46, "citationCount": 5558, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/889c81b4d7b7ed43a3f69f880ea60b0572e02e27", "doi": "10.1109/CVPR.2019.00075", "arxiv": "1902.09630" }, { "title": "Stacked Hourglass Networks for Human Pose Estimation", "authors": "Alejandro Newell, Kaiyu Yang, Jia Deng", "venue": "ECCV", "year": 2016, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "This work introduces a novel convolutional network architecture for the task of human pose estimation. Features are processed across all scales and consolidated to best capture the various spatial relationships associated with the body. We show how repeated bottom-up, top-down processing used in conjunction with intermediate supervision is critical to improving the performance of the network. We refer to the architecture as a “stacked hourglass” network based on the successive steps of pooling and upsampling that are done to produce a final set of predictions. State-of-the-art results are achieved on the FLIC and MPII benchmarks outcompeting all recent methods.", "motivation": "This work introduces a novel convolutional network architecture for the task of human pose estimation.", "implementation": "We show how repeated bottom-up, top-down processing used in conjunction with intermediate supervision is critical to improving the performance of the network.", "application": "State-of-the-art results are achieved on the FLIC and MPII benchmarks outcompeting all recent methods.", "problem": "This work introduces a novel convolutional network architecture for the task of human pose estimation.", "contribution": "We show how repeated bottom-up, top-down processing used in conjunction with intermediate supervision is critical to improving the performance of the network.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Stacked Hourglass Networks for Human Pose Estimation 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 5,443；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "ECCV", "Datasets / Benchmarks" ], "code": { "available": false, "url": "" }, "url": "https://www.semanticscholar.org/paper/848938e6199bad08f1db6f3239b260cfa901e95f", "updated": "2026-05-20", "topCitedRank": 47, "citationCount": 5443, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/848938e6199bad08f1db6f3239b260cfa901e95f", "doi": "10.1007/978-3-319-46484-8_29", "arxiv": "1603.06937" }, { "title": "ScanNet: Richly-Annotated 3D Reconstructions of Indoor Scenes", "authors": "Angela Dai, Angel X. Chang, M. Savva et al.", "venue": "CVPR", "year": 2017, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "A key requirement for leveraging supervised deep learning methods is the availability of large, labeled datasets. Unfortunately, in the context of RGB-D scene understanding, very little data is available – current datasets cover a small range of scene views and have limited semantic annotations. To address this issue, we introduce ScanNet, an RGB-D video dataset containing 2.5M views in 1513 scenes annotated with 3D camera poses, surface reconstructions, and semantic segmentations. To collect this data, we designed an easy-to-use and scalable RGB-D capture system that includes automated surface reconstruction and crowdsourced semantic annotation. We show that using this data helps achieve state-of-the-art performance on several 3D scene understanding tasks, including 3D object classification, semantic voxel labeling, and CAD model retrieval.", "motivation": "Unfortunately, in the context of RGB-D scene understanding, very little data is available – current datasets cover a small range of scene views and have limited semantic annotations.", "implementation": "To address this issue, we introduce ScanNet, an RGB-D video dataset containing 2.5M views in 1513 scenes annotated with 3D camera poses, surface reconstructions, and semantic segmentations.", "application": "To collect this data, we designed an easy-to-use and scalable RGB-D capture system that includes automated surface reconstruction and crowdsourced semantic annotation.", "problem": "Unfortunately, in the context of RGB-D scene understanding, very little data is available – current datasets cover a small range of scene views and have limited semantic annotations.", "contribution": "To address this issue, we introduce ScanNet, an RGB-D video dataset containing 2.5M views in 1513 scenes annotated with 3D camera poses, surface reconstructions, and semantic segmentations.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "ScanNet: Richly-Annotated 3D Reconstructions of Indoor Scenes 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 5,397；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Datasets / Benchmarks" ], "code": { "available": false, "url": "" }, "url": "https://arxiv.org/pdf/1702.04405", "updated": "2026-05-20", "topCitedRank": 48, "citationCount": 5397, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/e52e37cd91366f07df1f98e88f87010f494dd16e", "doi": "10.1109/CVPR.2017.261", "arxiv": "1702.04405" }, { "title": "Image Super-Resolution Using Very Deep Residual Channel Attention Networks", "authors": "Yulun Zhang, Kunpeng Li, Kai Li et al.", "venue": "ECCV", "year": 2018, "topic": "3D / Geometry / Driving", "modality": "3D / depth / optical flow / driving sensors", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "三维理解、几何重建、光流、深度或自动驾驶感知", "method": "几何约束、点云网络、连续场表示或多传感器数据集", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "Convolutional neural network (CNN) depth is of crucial importance for image super-resolution (SR). However, we observe that deeper networks for image SR are more difficult to train. The low-resolution inputs and features contain abundant low-frequency information, which is treated equally across channels, hence hindering the representational ability of CNNs. To solve these problems, we propose the very deep residual channel attention networks (RCAN). Specifically, we propose a residual in residual (RIR) structure to form very deep network, which consists of several residual groups with long skip connections. Each residual group contains some residual blocks with short skip connections. Meanwhile, RIR allows abundant low-frequency information to be bypassed through multiple skip connections, making the main network focus on learning high-frequency information. Furthermore, we propose a channel attention mechanism to adaptively rescale channel-wise features by considering interdependencies among channels. Extensive experiments show that our RCAN achieves better accuracy and visual improvements against state-of-the-art methods.", "motivation": "However, we observe that deeper networks for image SR are more difficult to train.", "implementation": "Convolutional neural network (CNN) depth is of crucial importance for image super-resolution (SR).", "application": "Specifically, we propose a residual in residual (RIR) structure to form very deep network, which consists of several residual groups with long skip connections.", "problem": "However, we observe that deeper networks for image SR are more difficult to train.", "contribution": "Convolutional neural network (CNN) depth is of crucial importance for image super-resolution (SR).", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Image Super-Resolution Using Very Deep Residual Channel Attention Networks 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 3D / Geometry / Driving 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 5,255；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何把真实传感噪声、动态场景和跨域泛化纳入可靠评测。", "directionNote": "该方向的高引用工作集中体现了 3D / Geometry / Driving 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "ECCV", "3D / Geometry / Driving" ], "code": { "available": false, "url": "" }, "url": "https://www.semanticscholar.org/paper/9775f8964a2eea1c9e35a02b1b906487396ea1f5", "updated": "2026-05-20", "topCitedRank": 49, "citationCount": 5255, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/9775f8964a2eea1c9e35a02b1b906487396ea1f5", "doi": "10.1007/978-3-030-01234-2_18", "arxiv": "1807.02758" }, { "title": "Arbitrary Style Transfer in Real-Time with Adaptive Instance Normalization", "authors": "Xun Huang, Serge J. Belongie", "venue": "ICCV", "year": 2017, "topic": "Generative Vision", "modality": "Image generation / editing", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "图像生成、编辑、翻译或高保真合成", "method": "GAN、diffusion、transformer 或条件生成模型", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "Gatys et al. recently introduced a neural algorithm that renders a content image in the style of another image, achieving so-called style transfer. However, their framework requires a slow iterative optimization process, which limits its practical application. Fast approximations with feed-forward neural networks have been proposed to speed up neural style transfer. Unfortunately, the speed improvement comes at a cost: the network is usually tied to a fixed set of styles and cannot adapt to arbitrary new styles. In this paper, we present a simple yet effective approach that for the first time enables arbitrary style transfer in real-time. At the heart of our method is a novel adaptive instance normalization (AdaIN) layer that aligns the mean and variance of the content features with those of the style features. Our method achieves speed comparable to the fastest existing approach, without the restriction to a pre-defined set of styles. In addition, our approach allows flexible user controls such as content-style trade-off, style interpolation, color & spatial controls, all using a single feed-forward neural network.", "motivation": "However, their framework requires a slow iterative optimization process, which limits its practical application.", "implementation": "recently introduced a neural algorithm that renders a content image in the style of another image, achieving so-called style transfer.", "application": "Unfortunately, the speed improvement comes at a cost: the network is usually tied to a fixed set of styles and cannot adapt to arbitrary new styles.", "problem": "However, their framework requires a slow iterative optimization process, which limits its practical application.", "contribution": "recently introduced a neural algorithm that renders a content image in the style of another image, achieving so-called style transfer.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Arbitrary Style Transfer in Real-Time with Adaptive Instance Normalization 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Generative Vision 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 5,207；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何平衡生成质量、可控性、版权风险和真实任务可靠性。", "directionNote": "该方向的高引用工作集中体现了 Generative Vision 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "ICCV", "Generative Vision" ], "code": { "available": false, "url": "" }, "url": "https://arxiv.org/pdf/1703.06868", "updated": "2026-05-20", "topCitedRank": 50, "citationCount": 5207, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/be0ef77fb0345c5851bb5d297f3ed84ae3c581ee", "doi": "10.1109/ICCV.2017.167", "arxiv": "1703.06868" }, { "title": "Adversarial Discriminative Domain Adaptation", "authors": "E. Tzeng, Judy Hoffman, Kate Saenko et al.", "venue": "CVPR", "year": 2017, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "Adversarial learning methods are a promising approach to training robust deep networks, and can generate complex samples across diverse domains. They can also improve recognition despite the presence of domain shift or dataset bias: recent adversarial approaches to unsupervised domain adaptation reduce the difference between the training and test domain distributions and thus improve generalization performance. However, while generative adversarial networks (GANs) show compelling visualizations, they are not optimal on discriminative tasks and can be limited to smaller shifts. On the other hand, discriminative approaches can handle larger domain shifts, but impose tied weights on the model and do not exploit a GAN-based loss. In this work, we first outline a novel generalized framework for adversarial adaptation, which subsumes recent state-of-the-art approaches as special cases, and use this generalized view to better relate prior approaches. We then propose a previously unexplored instance of our general framework which combines discriminative modeling, untied weight sharing, and a GAN loss, which we call Adversarial Discriminative Domain Adaptation (ADDA). We show that ADDA is more effective yet considerably simpler than competing domain-adversarial methods, and demonstrate the promise of our approach by exceeding state-of-the-art unsupervised adaptation results on standard domain adaptation tasks as well as a difficult cross-modality object classification task.", "motivation": "They can also improve recognition despite the presence of domain shift or dataset bias: recent adversarial approaches to unsupervised domain adaptation reduce the difference between the training and test domain distributions and thus improve generalization performance.", "implementation": "On the other hand, discriminative approaches can handle larger domain shifts, but impose tied weights on the model and do not exploit a GAN-based loss.", "application": "We show that ADDA is more effective yet considerably simpler than competing domain-adversarial methods, and demonstrate the promise of our approach by exceeding state-of-the-art unsupervised adaptation results on standard domain adaptation tasks as well as a difficult cross-modality object classification task.", "problem": "They can also improve recognition despite the presence of domain shift or dataset bias: recent adversarial approaches to unsupervised domain adaptation reduce the difference between the training and test domain distributions and thus improve generalization performance.", "contribution": "On the other hand, discriminative approaches can handle larger domain shifts, but impose tied weights on the model and do not exploit a GAN-based loss.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Adversarial Discriminative Domain Adaptation 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 5,191；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Datasets / Benchmarks" ], "code": { "available": false, "url": "" }, "url": "https://arxiv.org/pdf/1702.05464", "updated": "2026-05-20", "topCitedRank": 51, "citationCount": 5191, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/345afa0e85cb2f5cb438ae44027499ff2c392409", "doi": "10.1109/CVPR.2017.316", "arxiv": "1702.05464" }, { "title": "Improved Baselines with Visual Instruction Tuning", "authors": "Haotian Liu, Chunyuan Li, Yuheng Li et al.", "venue": "CVPR", "year": 2023, "topic": "Vision-Language / Foundation Models", "modality": "Image-text / foundation model", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "视觉-语言理解、开放词表识别或基础模型适配", "method": "大规模预训练、多模态对齐或提示式视觉建模", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "Large multimodal models (LMM) have recently shown encouraging progress with visual instruction tuning. In this paper, we present the first systematic study to investigate the design choices of LMMs in a controlled setting under the LLaVA framework. We show that the fully-connected vision-language connector in LLaVA is surprisingly power-ful and data-efficient. With simple modifications to LLa VA, namely, using CLIP- ViT-L-336px with an MLP projection and adding academic-task-oriented VQA data with response formatting prompts, we establish stronger baselines that achieve state-of-the-art across 11 benchmarks. Our final 13B checkpoint uses merely 1.2M publicly available data, and finishes full training in ~ 1 day on a single 8-AI00 node. Furthermore, we present some early exploration of open problems in LMMs, including scaling to higher resolution inputs, compositional capabilities, and model hallucination, etc. We hope this makes state-of-the-art LMM research more accessible. Code and model will be publicly available.", "motivation": "Large multimodal models (LMM) have recently shown encouraging progress with visual instruction tuning.", "implementation": "In this paper, we present the first systematic study to investigate the design choices of LMMs in a controlled setting under the LLaVA framework.", "application": "Our final 13B checkpoint uses merely 1.2M publicly available data, and finishes full training in ~ 1 day on a single 8-AI00 node.", "problem": "Large multimodal models (LMM) have recently shown encouraging progress with visual instruction tuning.", "contribution": "In this paper, we present the first systematic study to investigate the design choices of LMMs in a controlled setting under the LLaVA framework.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Improved Baselines with Visual Instruction Tuning 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Vision-Language / Foundation Models 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 5,067；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何降低标注和算力成本，同时保持可控性、可解释性和领域泛化。", "directionNote": "该方向的高引用工作集中体现了 Vision-Language / Foundation Models 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Vision-Language / Foundation Models" ], "code": { "available": false, "url": "" }, "url": "https://arxiv.org/pdf/2310.03744", "updated": "2026-05-20", "topCitedRank": 52, "citationCount": 5067, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/124d4d374fbef2016fa9880489871a58a7450644", "doi": "10.1109/CVPR52733.2024.02484", "arxiv": "2310.03744" }, { "title": "Deep High-Resolution Representation Learning for Human Pose Estimation", "authors": "Ke Sun, Bin Xiao, Dong Liu et al.", "venue": "CVPR", "year": 2019, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "In this paper, we are interested in the human pose estimation problem with a focus on learning reliable high-resolution representations. Most existing methods recover high-resolution representations from low-resolution representations produced by a high-to-low resolution network. Instead, our proposed network maintains high-resolution representations through the whole process. We start from a high-resolution subnetwork as the first stage, gradually add high-to-low resolution subnetworks one by one to form more stages, and connect the mutli-resolution subnetworks in parallel. We conduct repeated multi-scale fusions such that each of the high-to-low resolution representations receives information from other parallel representations over and over, leading to rich high-resolution representations. As a result, the predicted keypoint heatmap is potentially more accurate and spatially more precise. We empirically demonstrate the effectiveness of our network through the superior pose estimation results over two benchmark datasets: the COCO keypoint detection dataset and the MPII Human Pose dataset. In addition, we show the superiority of our network in pose tracking on the PoseTrack dataset. The code and models have been publicly available at https://github.com/leoxiaobin/deep-high-resolution-net.pytorch.", "motivation": "In this paper, we are interested in the human pose estimation problem with a focus on learning reliable high-resolution representations.", "implementation": "Most existing methods recover high-resolution representations from low-resolution representations produced by a high-to-low resolution network.", "application": "We empirically demonstrate the effectiveness of our network through the superior pose estimation results over two benchmark datasets: the COCO keypoint detection dataset and the MPII Human Pose dataset.", "problem": "In this paper, we are interested in the human pose estimation problem with a focus on learning reliable high-resolution representations.", "contribution": "Most existing methods recover high-resolution representations from low-resolution representations produced by a high-to-low resolution network.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Deep High-Resolution Representation Learning for Human Pose Estimation 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 5,044；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Datasets / Benchmarks" ], "code": { "available": true, "url": "https://github.com/leoxiaobin/deep-high-resolution-net.pytorch" }, "url": "https://arxiv.org/pdf/1902.09212", "updated": "2026-05-20", "topCitedRank": 53, "citationCount": 5044, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/6303bac53abd725c3b458190a6abe389a4a1e72d", "doi": "10.1109/CVPR.2019.00584", "arxiv": "1902.09212" }, { "title": "Least Squares Generative Adversarial Networks", "authors": "Xudong Mao, Qing Li, Haoran Xie et al.", "venue": "ICCV", "year": 2016, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "Unsupervised learning with generative adversarial networks (GANs) has proven hugely successful. Regular GANs hypothesize the discriminator as a classifier with the sigmoid cross entropy loss function. However, we found that this loss function may lead to the vanishing gradients problem during the learning process. To overcome such a problem, we propose in this paper the Least Squares Generative Adversarial Networks (LSGANs) which adopt the least squares loss function for the discriminator. We show that minimizing the objective function of LSGAN yields minimizing the Pearson X2 divergence. There are two benefits of LSGANs over regular GANs. First, LSGANs are able to generate higher quality images than regular GANs. Second, LSGANs perform more stable during the learning process. We evaluate LSGANs on LSUN and CIFAR-10 datasets and the experimental results show that the images generated by LSGANs are of better quality than the ones generated by regular GANs. We also conduct two comparison experiments between LSGANs and regular GANs to illustrate the stability of LSGANs.", "motivation": "However, we found that this loss function may lead to the vanishing gradients problem during the learning process.", "implementation": "Regular GANs hypothesize the discriminator as a classifier with the sigmoid cross entropy loss function.", "application": "We evaluate LSGANs on LSUN and CIFAR-10 datasets and the experimental results show that the images generated by LSGANs are of better quality than the ones generated by regular GANs.", "problem": "However, we found that this loss function may lead to the vanishing gradients problem during the learning process.", "contribution": "Regular GANs hypothesize the discriminator as a classifier with the sigmoid cross entropy loss function.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Least Squares Generative Adversarial Networks 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 5,024；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "ICCV", "Datasets / Benchmarks" ], "code": { "available": false, "url": "" }, "url": "https://arxiv.org/pdf/1611.04076", "updated": "2026-05-20", "topCitedRank": 54, "citationCount": 5024, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/74ff6d48f9c62e937023106629d27ef2d2ddf8bc", "doi": "10.1109/ICCV.2017.304", "arxiv": "1611.04076" }, { "title": "Exploring Simple Siamese Representation Learning", "authors": "Xinlei Chen, Kaiming He", "venue": "CVPR", "year": 2020, "topic": "Representation Learning", "modality": "Visual representation", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "自监督、半监督或可迁移视觉表征学习", "method": "对比学习、掩码建模、数据增强或持续学习", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "Siamese networks have become a common structure in various recent models for unsupervised visual representation learning. These models maximize the similarity between two augmentations of one image, subject to certain conditions for avoiding collapsing solutions. In this paper, we report surprising empirical results that simple Siamese networks can learn meaningful representations even using none of the following: (i) negative sample pairs, (ii) large batches, (iii) momentum encoders. Our experiments show that collapsing solutions do exist for the loss and structure, but a stop-gradient operation plays an essential role in preventing collapsing. We provide a hypothesis on the implication of stop-gradient, and further show proof-of-concept experiments verifying it. Our \"SimSiam\" method achieves competitive results on ImageNet and downstream tasks. We hope this simple baseline will motivate people to rethink the roles of Siamese architectures for unsupervised representation learning. Code is made available.1", "motivation": "Siamese networks have become a common structure in various recent models for unsupervised visual representation learning.", "implementation": "Our experiments show that collapsing solutions do exist for the loss and structure, but a stop-gradient operation plays an essential role in preventing collapsing.", "application": "Our \"SimSiam\" method achieves competitive results on ImageNet and downstream tasks.", "problem": "Siamese networks have become a common structure in various recent models for unsupervised visual representation learning.", "contribution": "Our experiments show that collapsing solutions do exist for the loss and structure, but a stop-gradient operation plays an essential role in preventing collapsing.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Exploring Simple Siamese Representation Learning 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Representation Learning 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 4,959；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何评估表征的因果性、可迁移性和下游任务稳定性。", "directionNote": "该方向的高引用工作集中体现了 Representation Learning 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Representation Learning", "award-winning", "Honorable Mention" ], "code": { "available": false, "url": "" }, "url": "https://arxiv.org/pdf/2011.10566", "updated": "2026-05-20", "topCitedRank": 55, "citationCount": 4959, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/0e23d2f14e7e56e81538f4a63e11689d8ac1eb9d", "doi": "10.1109/CVPR46437.2021.01549", "arxiv": "2011.10566", "awards": [ { "conference": "CVPR", "year": 2021, "award": "Honorable Mention", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ] }, { "title": "Pyramid Vision Transformer: A Versatile Backbone for Dense Prediction without Convolutions", "authors": "Wenhai Wang, Enze Xie, Xiang Li et al.", "venue": "ICCV", "year": 2021, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "Although convolutional neural networks (CNNs) have achieved great success in computer vision, this work investigates a simpler, convolution-free backbone network use-fid for many dense prediction tasks. Unlike the recently-proposed Vision Transformer (ViT) that was designed for image classification specifically, we introduce the Pyramid Vision Transformer (PVT), which overcomes the difficulties of porting Transformer to various dense prediction tasks. PVT has several merits compared to current state of the arts. (1) Different from ViT that typically yields low-resolution outputs and incurs high computational and memory costs, PVT not only can be trained on dense partitions of an image to achieve high output resolution, which is important for dense prediction, but also uses a progressive shrinking pyramid to reduce the computations of large feature maps. (2) PVT inherits the advantages of both CNN and Transformer, making it a unified backbone for various vision tasks without convolutions, where it can be used as a direct replacement for CNN backbones. (3) We validate PVT through extensive experiments, showing that it boosts the performance of many downstream tasks, including object detection, instance and semantic segmentation. For example, with a comparable number of parameters, PVT+RetinaNet achieves 40.4 AP on the COCO dataset, surpassing ResNet50+RetinNet (36.3 AP) by 4.1 absolute AP (see Figure 2). We hope that PVT could, serre as an alternative and useful backbone for pixel-level predictions and facilitate future research.", "motivation": "Although convolutional neural networks (CNNs) have achieved great success in computer vision, this work investigates a simpler, convolution-free backbone network use-fid for many dense prediction tasks.", "implementation": "Unlike the recently-proposed Vision Transformer (ViT) that was designed for image classification specifically, we introduce the Pyramid Vision Transformer (PVT), which overcomes the difficulties of porting Transformer to various dense prediction tasks.", "application": "(2) PVT inherits the advantages of both CNN and Transformer, making it a unified backbone for various vision tasks without convolutions, where it can be used as a direct replacement for CNN backbones.", "problem": "Although convolutional neural networks (CNNs) have achieved great success in computer vision, this work investigates a simpler, convolution-free backbone network use-fid for many dense prediction tasks.", "contribution": "Unlike the recently-proposed Vision Transformer (ViT) that was designed for image classification specifically, we introduce the Pyramid Vision Transformer (PVT), which overcomes the difficulties of porting Transformer to various dense prediction tasks.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Pyramid Vision Transformer: A Versatile Backbone for Dense Prediction without Convolutions 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 4,953；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "ICCV", "Datasets / Benchmarks" ], "code": { "available": false, "url": "" }, "url": "https://arxiv.org/pdf/2102.12122", "updated": "2026-05-20", "topCitedRank": 56, "citationCount": 4953, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/8fb1c04dab87ca6c116495e4d03c46c9547e4ec3", "doi": "10.1109/ICCV48922.2021.00061", "arxiv": "2102.12122" }, { "title": "iCaRL: Incremental Classifier and Representation Learning", "authors": "Sylvestre-Alvise Rebuffi, Alexander Kolesnikov, G. Sperl et al.", "venue": "CVPR", "year": 2016, "topic": "Representation Learning", "modality": "Visual representation", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "自监督、半监督或可迁移视觉表征学习", "method": "对比学习、掩码建模、数据增强或持续学习", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "A major open problem on the road to artificial intelligence is the development of incrementally learning systems that learn about more and more concepts over time from a stream of data. In this work, we introduce a new training strategy, iCaRL, that allows learning in such a class-incremental way: only the training data for a small number of classes has to be present at the same time and new classes can be added progressively. iCaRL learns strong classifiers and a data representation simultaneously. This distinguishes it from earlier works that were fundamentally limited to fixed data representations and therefore incompatible with deep learning architectures. We show by experiments on CIFAR-100 and ImageNet ILSVRC 2012 data that iCaRL can learn many classes incrementally over a long period of time where other strategies quickly fail.", "motivation": "A major open problem on the road to artificial intelligence is the development of incrementally learning systems that learn about more and more concepts over time from a stream of data.", "implementation": "In this work, we introduce a new training strategy, iCaRL, that allows learning in such a class-incremental way: only the training data for a small number of classes has to be present at the same time and new classes can be added progressively.", "application": "We show by experiments on CIFAR-100 and ImageNet ILSVRC 2012 data that iCaRL can learn many classes incrementally over a long period of time where other strategies quickly fail.", "problem": "A major open problem on the road to artificial intelligence is the development of incrementally learning systems that learn about more and more concepts over time from a stream of data.", "contribution": "In this work, we introduce a new training strategy, iCaRL, that allows learning in such a class-incremental way: only the training data for a small number of classes has to be present at the same time and new classes can be added progressively.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "iCaRL: Incremental Classifier and Representation Learning 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Representation Learning 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 4,725；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何评估表征的因果性、可迁移性和下游任务稳定性。", "directionNote": "该方向的高引用工作集中体现了 Representation Learning 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Representation Learning" ], "code": { "available": false, "url": "" }, "url": "https://arxiv.org/pdf/1611.07725", "updated": "2026-05-20", "topCitedRank": 57, "citationCount": 4725, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/1703631a938b397ba7e858161ce16448f6046d6f", "doi": "10.1109/CVPR.2017.587", "arxiv": "1611.07725" }, { "title": "Coordinate Attention for Efficient Mobile Network Design", "authors": "Qibin Hou, Daquan Zhou, Jiashi Feng", "venue": "CVPR", "year": 2021, "topic": "Object Detection", "modality": "RGB / multi-scale visual features", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "目标检测、开放集检测或检测损失设计", "method": "检测框架、特征金字塔、transformer detector 或检测损失", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "Recent studies on mobile network design have demonstrated the remarkable effectiveness of channel attention (e.g., the Squeeze-and-Excitation attention) for lifting model performance, but they generally neglect the positional information, which is important for generating spatially selective attention maps. In this paper, we propose a novel attention mechanism for mobile networks by embedding positional information into channel attention, which we call \"coordinate attention\". Unlike channel attention that transforms a feature tensor to a single feature vector via 2D global pooling, the coordinate attention factorizes channel attention into two 1D feature encoding processes that aggregate features along the two spatial directions, respectively. In this way, long-range dependencies can be captured along one spatial direction and meanwhile precise positional information can be preserved along the other spatial direction. The resulting feature maps are then encoded separately into a pair of direction-aware and position-sensitive attention maps that can be complementarily applied to the input feature map to augment the representations of the objects of interest. Our coordinate attention is simple and can be flexibly plugged into classic mobile networks, such as MobileNetV2, MobileNeXt, and EfficientNet with nearly no computational overhead. Extensive experiments demonstrate that our coordinate attention is not only beneficial to ImageNet classification but more interestingly, behaves better in down-stream tasks, such as object detection and semantic segmentation. Code is available at https://github.com/Andrew-Qibin/CoordAttention.", "motivation": "Recent studies on mobile network design have demonstrated the remarkable effectiveness of channel attention (e.g., the Squeeze-and-Excitation attention) for lifting model performance, but they generally neglect the positional information, which is important for generating spatially selective attention maps.", "implementation": "In this paper, we propose a novel attention mechanism for mobile networks by embedding positional information into channel attention, which we call \"coordinate attention\".", "application": "Extensive experiments demonstrate that our coordinate attention is not only beneficial to ImageNet classification but more interestingly, behaves better in down-stream tasks, such as object detection and semantic segmentation.", "problem": "Recent studies on mobile network design have demonstrated the remarkable effectiveness of channel attention (e.g., the Squeeze-and-Excitation attention) for lifting model performance, but they generally neglect the positional information, which is important for generating spatially selective attention maps.", "contribution": "In this paper, we propose a novel attention mechanism for mobile networks by embedding positional information into channel attention, which we call \"coordinate attention\".", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Coordinate Attention for Efficient Mobile Network Design 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Object Detection 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 4,709；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何在开放世界、长尾类别和实时部署中保持稳定检测性能。", "directionNote": "该方向的高引用工作集中体现了 Object Detection 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Object Detection" ], "code": { "available": true, "url": "https://github.com/Andrew-Qibin/CoordAttention" }, "url": "https://arxiv.org/pdf/2103.02907", "updated": "2026-05-20", "topCitedRank": 58, "citationCount": 4709, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/70cf7c785952375e8061c92235aa20e94b02ecd4", "doi": "10.1109/CVPR46437.2021.01350", "arxiv": "2103.02907" }, { "title": "XNOR-Net: ImageNet Classification Using Binary Convolutional Neural Networks", "authors": "Mohammad Rastegari, Vicente Ordonez, J. Redmon et al.", "venue": "ECCV", "year": 2016, "topic": "Video / Human Understanding", "modality": "Video / human-centric data", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "视频理解、动作识别、人体姿态或轨迹预测", "method": "时序建模、人体结构先验或视频 transformer", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "We propose two efficient approximations to standard convolutional neural networks: Binary-Weight-Networks and XNOR-Networks. In Binary-Weight-Networks, the filters are approximated with binary values resulting in 32\\(\\times \\) memory saving. In XNOR-Networks, both the filters and the input to convolutional layers are binary. XNOR-Networks approximate convolutions using primarily binary operations. This results in 58\\(\\times \\) faster convolutional operations (in terms of number of the high precision operations) and 32\\(\\times \\) memory savings. XNOR-Nets offer the possibility of running state-of-the-art networks on CPUs (rather than GPUs) in real-time. Our binary networks are simple, accurate, efficient, and work on challenging visual tasks. We evaluate our approach on the ImageNet classification task. The classification accuracy with a Binary-Weight-Network version of AlexNet is the same as the full-precision AlexNet. We compare our method with recent network binarization methods, BinaryConnect and BinaryNets, and outperform these methods by large margins on ImageNet, more than \\(16\\,\\%\\) in top-1 accuracy. Our code is available at: http://allenai.org/plato/xnornet.", "motivation": "Our binary networks are simple, accurate, efficient, and work on challenging visual tasks.", "implementation": "We propose two efficient approximations to standard convolutional neural networks: Binary-Weight-Networks and XNOR-Networks.", "application": "We evaluate our approach on the ImageNet classification task.", "problem": "Our binary networks are simple, accurate, efficient, and work on challenging visual tasks.", "contribution": "We propose two efficient approximations to standard convolutional neural networks: Binary-Weight-Networks and XNOR-Networks.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "XNOR-Net: ImageNet Classification Using Binary Convolutional Neural Networks 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Video / Human Understanding 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 4,706；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何在长时序、遮挡和复杂交互下保持鲁棒理解。", "directionNote": "该方向的高引用工作集中体现了 Video / Human Understanding 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "ECCV", "Video / Human Understanding" ], "code": { "available": false, "url": "" }, "url": "https://www.semanticscholar.org/paper/b649a98ce77ece8cd7638bb74ab77d22d9be77e7", "updated": "2026-05-20", "topCitedRank": 59, "citationCount": 4706, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/b649a98ce77ece8cd7638bb74ab77d22d9be77e7", "doi": "10.1007/978-3-319-46493-0_32", "arxiv": "1603.05279" }, { "title": "DeepSDF: Learning Continuous Signed Distance Functions for Shape Representation", "authors": "J. Park, Peter R. Florence, Julian Straub et al.", "venue": "CVPR", "year": 2019, "topic": "3D / Geometry / Driving", "modality": "3D / depth / optical flow / driving sensors", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "三维理解、几何重建、光流、深度或自动驾驶感知", "method": "几何约束、点云网络、连续场表示或多传感器数据集", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "Computer graphics, 3D computer vision and robotics communities have produced multiple approaches to representing 3D geometry for rendering and reconstruction. These provide trade-offs across fidelity, efficiency and compression capabilities. In this work, we introduce DeepSDF, a learned continuous Signed Distance Function (SDF) representation of a class of shapes that enables high quality shape representation, interpolation and completion from partial and noisy 3D input data. DeepSDF, like its classical counterpart, represents a shape's surface by a continuous volumetric field: the magnitude of a point in the field represents the distance to the surface boundary and the sign indicates whether the region is inside (-) or outside (+) of the shape, hence our representation implicitly encodes a shape's boundary as the zero-level-set of the learned function while explicitly representing the classification of space as being part of the shapes interior or not. While classical SDF's both in analytical or discretized voxel form typically represent the surface of a single shape, DeepSDF can represent an entire class of shapes. Furthermore, we show state-of-the-art performance for learned 3D shape representation and completion while reducing the model size by an order of magnitude compared with previous work.", "motivation": "Computer graphics, 3D computer vision and robotics communities have produced multiple approaches to representing 3D geometry for rendering and reconstruction.", "implementation": "In this work, we introduce DeepSDF, a learned continuous Signed Distance Function (SDF) representation of a class of shapes that enables high quality shape representation, interpolation and completion from partial and noisy 3D input data.", "application": "DeepSDF, like its classical counterpart, represents a shape's surface by a continuous volumetric field: the magnitude of a point in the field represents the distance to the surface boundary and the sign indicates whether the region is inside (-) or outside (+) of the shape, hence our representation implicitly encodes a shape's boundary as the zero-level-set…", "problem": "Computer graphics, 3D computer vision and robotics communities have produced multiple approaches to representing 3D geometry for rendering and reconstruction.", "contribution": "In this work, we introduce DeepSDF, a learned continuous Signed Distance Function (SDF) representation of a class of shapes that enables high quality shape representation, interpolation and completion from partial and noisy 3D input data.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "DeepSDF: Learning Continuous Signed Distance Functions for Shape Representation 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 3D / Geometry / Driving 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 4,492；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何把真实传感噪声、动态场景和跨域泛化纳入可靠评测。", "directionNote": "该方向的高引用工作集中体现了 3D / Geometry / Driving 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "3D / Geometry / Driving" ], "code": { "available": false, "url": "" }, "url": "https://www.semanticscholar.org/paper/dd81523b9accdf1c13cd37f76b22ab27d84b7a42", "updated": "2026-05-20", "topCitedRank": 60, "citationCount": 4492, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/dd81523b9accdf1c13cd37f76b22ab27d84b7a42", "doi": "10.1109/CVPR.2019.00025", "arxiv": "1901.05103" }, { "title": "PointPillars: Fast Encoders for Object Detection From Point Clouds", "authors": "Alex H. Lang, Sourabh Vora, Holger Caesar et al.", "venue": "CVPR", "year": 2018, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "Object detection in point clouds is an important aspect of many robotics applications such as autonomous driving. In this paper, we consider the problem of encoding a point cloud into a format appropriate for a downstream detection pipeline. Recent literature suggests two types of encoders; fixed encoders tend to be fast but sacrifice accuracy, while encoders that are learned from data are more accurate, but slower. In this work, we propose PointPillars, a novel encoder which utilizes PointNets to learn a representation of point clouds organized in vertical columns (pillars). While the encoded features can be used with any standard 2D convolutional detection architecture, we further propose a lean downstream network. Extensive experimentation shows that PointPillars outperforms previous encoders with respect to both speed and accuracy by a large margin. Despite only using lidar, our full detection pipeline significantly outperforms the state of the art, even among fusion methods, with respect to both the 3D and bird’s eye view KITTI benchmarks. This detection performance is achieved while running at 62 Hz: a 2 - 4 fold runtime improvement. A faster version of our method matches the state of the art at 105 Hz. These benchmarks suggest that PointPillars is an appropriate encoding for object detection in point clouds.", "motivation": "In this paper, we consider the problem of encoding a point cloud into a format appropriate for a downstream detection pipeline.", "implementation": "In this work, we propose PointPillars, a novel encoder which utilizes PointNets to learn a representation of point clouds organized in vertical columns (pillars).", "application": "Object detection in point clouds is an important aspect of many robotics applications such as autonomous driving.", "problem": "In this paper, we consider the problem of encoding a point cloud into a format appropriate for a downstream detection pipeline.", "contribution": "In this work, we propose PointPillars, a novel encoder which utilizes PointNets to learn a representation of point clouds organized in vertical columns (pillars).", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "PointPillars: Fast Encoders for Object Detection From Point Clouds 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 4,446；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Datasets / Benchmarks" ], "code": { "available": false, "url": "" }, "url": "https://arxiv.org/pdf/1812.05784", "updated": "2026-05-20", "topCitedRank": 61, "citationCount": 4446, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/3bb322718d64a34b91b29c8230c5978de5d7fb7a", "doi": "10.1109/CVPR.2019.01298", "arxiv": "1812.05784" }, { "title": "Temporal Segment Networks: Towards Good Practices for Deep Action Recognition", "authors": "Limin Wang, Yuanjun Xiong, Zhe Wang et al.", "venue": "ECCV", "year": 2016, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "Deep convolutional networks have achieved great success for visual recognition in still images. However, for action recognition in videos, the advantage over traditional methods is not so evident. This paper aims to discover the principles to design effective ConvNet architectures for action recognition in videos and learn these models given limited training samples. Our first contribution is temporal segment network (TSN), a novel framework for video-based action recognition. which is based on the idea of long-range temporal structure modeling. It combines a sparse temporal sampling strategy and video-level supervision to enable efficient and effective learning using the whole action video. The other contribution is our study on a series of good practices in learning ConvNets on video data with the help of temporal segment network. Our approach obtains the state-the-of-art performance on the datasets of HMDB51 (\\( 69.4\\,\\% \\)) and UCF101 (\\( 94.2\\,\\% \\)). We also visualize the learned ConvNet models, which qualitatively demonstrates the effectiveness of temporal segment network and the proposed good practices (Models and code at https://github.com/yjxiong/temporal-segment-networks).", "motivation": "However, for action recognition in videos, the advantage over traditional methods is not so evident.", "implementation": "Our first contribution is temporal segment network (TSN), a novel framework for video-based action recognition.", "application": "Deep convolutional networks have achieved great success for visual recognition in still images.", "problem": "However, for action recognition in videos, the advantage over traditional methods is not so evident.", "contribution": "Our first contribution is temporal segment network (TSN), a novel framework for video-based action recognition.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Temporal Segment Networks: Towards Good Practices for Deep Action Recognition 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 4,245；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "ECCV", "Datasets / Benchmarks" ], "code": { "available": true, "url": "https://github.com/yjxiong/temporal-segment-networks" }, "url": "https://www.semanticscholar.org/paper/ea3d7de6c0880e14455b9acb28f1bc1234321456", "updated": "2026-05-20", "topCitedRank": 62, "citationCount": 4245, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/ea3d7de6c0880e14455b9acb28f1bc1234321456", "doi": "10.1007/978-3-319-46484-8_2", "arxiv": "1608.00859" }, { "title": "Taming Transformers for High-Resolution Image Synthesis", "authors": "Patrick Esser, Robin Rombach, B. Ommer", "venue": "CVPR", "year": 2020, "topic": "Generative Vision", "modality": "Image generation / editing", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "图像生成、编辑、翻译或高保真合成", "method": "GAN、diffusion、transformer 或条件生成模型", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "Designed to learn long-range interactions on sequential data, transformers continue to show state-of-the-art results on a wide variety of tasks. In contrast to CNNs, they contain no inductive bias that prioritizes local interactions. This makes them expressive, but also computationally infeasible for long sequences, such as high-resolution images. We demonstrate how combining the effectiveness of the inductive bias of CNNs with the expressivity of transformers enables them to model and thereby synthesize high-resolution images. We show how to (i) use CNNs to learn a context-rich vocabulary of image constituents, and in turn (ii) utilize transformers to efficiently model their composition within high-resolution images. Our approach is readily applied to conditional synthesis tasks, where both non-spatial information, such as object classes, and spatial information, such as segmentations, can control the generated image. In particular, we present the first results on semantically-guided synthesis of megapixel images with transformers. Project page at https://git.io/JLlvY.", "motivation": "Designed to learn long-range interactions on sequential data, transformers continue to show state-of-the-art results on a wide variety of tasks.", "implementation": "We demonstrate how combining the effectiveness of the inductive bias of CNNs with the expressivity of transformers enables them to model and thereby synthesize high-resolution images.", "application": "Our approach is readily applied to conditional synthesis tasks, where both non-spatial information, such as object classes, and spatial information, such as segmentations, can control the generated image.", "problem": "Designed to learn long-range interactions on sequential data, transformers continue to show state-of-the-art results on a wide variety of tasks.", "contribution": "We demonstrate how combining the effectiveness of the inductive bias of CNNs with the expressivity of transformers enables them to model and thereby synthesize high-resolution images.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Taming Transformers for High-Resolution Image Synthesis 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Generative Vision 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 4,222；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何平衡生成质量、可控性、版权风险和真实任务可靠性。", "directionNote": "该方向的高引用工作集中体现了 Generative Vision 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Generative Vision" ], "code": { "available": false, "url": "" }, "url": "https://arxiv.org/pdf/2012.09841", "updated": "2026-05-20", "topCitedRank": 63, "citationCount": 4222, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/47f7ec3d0a5e6e83b6768ece35206a94dc81919c", "doi": "10.1109/CVPR46437.2021.01268", "arxiv": "2012.09841" }, { "title": "DreamBooth: Fine Tuning Text-to-Image Diffusion Models for Subject-Driven Generation", "authors": "Nataniel Ruiz, Yuanzhen Li, Varun Jampani et al.", "venue": "CVPR", "year": 2022, "topic": "Vision-Language / Foundation Models", "modality": "Image-text / foundation model", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "视觉-语言理解、开放词表识别或基础模型适配", "method": "大规模预训练、多模态对齐或提示式视觉建模", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "Large text-to-image models achieved a remarkable leap in the evolution of AI, enabling high-quality and diverse synthesis of images from a given text prompt. However, these models lack the ability to mimic the appearance of subjects in a given reference set and synthesize novel renditions of them in different contexts. In this work, we present a new approach for “personalization” of text-to-image diffusion models. Given as input just a few images of a subject, we fine-tune a pretrained text-to-image model such that it learns to bind a unique identifier with that specific subject. Once the subject is embedded in the output domain of the model, the unique identifier can be used to synthesize novel photorealistic images of the subject contextualized in different scenes. By leveraging the semantic prior embedded in the model with a new autogenous class-specific prior preservation loss, our technique enables synthesizing the subject in diverse scenes, poses, views and lighting conditions that do not appear in the reference images. We apply our technique to several previously-unassailable tasks, including subject recontextualization, text-guided view synthesis, and artistic rendering, all while preserving the subject's key features. We also provide a new dataset and evaluation protocol for this new task of subject-driven generation. Project page: https://dreambooth.github.io/", "motivation": "However, these models lack the ability to mimic the appearance of subjects in a given reference set and synthesize novel renditions of them in different contexts.", "implementation": "Given as input just a few images of a subject, we fine-tune a pretrained text-to-image model such that it learns to bind a unique identifier with that specific subject.", "application": "Large text-to-image models achieved a remarkable leap in the evolution of AI, enabling high-quality and diverse synthesis of images from a given text prompt.", "problem": "However, these models lack the ability to mimic the appearance of subjects in a given reference set and synthesize novel renditions of them in different contexts.", "contribution": "Given as input just a few images of a subject, we fine-tune a pretrained text-to-image model such that it learns to bind a unique identifier with that specific subject.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "DreamBooth: Fine Tuning Text-to-Image Diffusion Models for Subject-Driven Generation 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Vision-Language / Foundation Models 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 4,174；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何降低标注和算力成本，同时保持可控性、可解释性和领域泛化。", "directionNote": "该方向的高引用工作集中体现了 Vision-Language / Foundation Models 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Vision-Language / Foundation Models", "award-winning", "Honorable Mention" ], "code": { "available": false, "url": "" }, "url": "https://arxiv.org/pdf/2208.12242", "updated": "2026-05-20", "topCitedRank": 64, "citationCount": 4174, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/5b19bf6c3f4b25cac96362c98b930cf4b37f6744", "doi": "10.1109/CVPR52729.2023.02155", "arxiv": "2208.12242", "awards": [ { "conference": "CVPR", "year": 2023, "award": "Honorable Mention", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ] }, { "title": "SlowFast Networks for Video Recognition", "authors": "Christoph Feichtenhofer, Haoqi Fan, Jitendra Malik et al.", "venue": "ICCV", "year": 2018, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "We present SlowFast networks for video recognition. Our model involves (i) a Slow pathway, operating at low frame rate, to capture spatial semantics, and (ii) a Fast pathway, operating at high frame rate, to capture motion at fine temporal resolution. The Fast pathway can be made very lightweight by reducing its channel capacity, yet can learn useful temporal information for video recognition. Our models achieve strong performance for both action classification and detection in video, and large improvements are pin-pointed as contributions by our SlowFast concept. We report state-of-the-art accuracy on major video recognition benchmarks, Kinetics, Charades and AVA. Code has been made available at: https://github.com/facebookresearch/SlowFast.", "motivation": "We present SlowFast networks for video recognition.", "implementation": "Our model involves (i) a Slow pathway, operating at low frame rate, to capture spatial semantics, and (ii) a Fast pathway, operating at high frame rate, to capture motion at fine temporal resolution.", "application": "The Fast pathway can be made very lightweight by reducing its channel capacity, yet can learn useful temporal information for video recognition.", "problem": "We present SlowFast networks for video recognition.", "contribution": "Our model involves (i) a Slow pathway, operating at low frame rate, to capture spatial semantics, and (ii) a Fast pathway, operating at high frame rate, to capture motion at fine temporal resolution.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "SlowFast Networks for Video Recognition 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 4,130；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "ICCV", "Datasets / Benchmarks" ], "code": { "available": true, "url": "https://github.com/facebookresearch/SlowFast" }, "url": "https://arxiv.org/pdf/1812.03982", "updated": "2026-05-20", "topCitedRank": 65, "citationCount": 4130, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/8b47b9c3c35b2b2a78bff7822605b3040f87d699", "doi": "10.1109/ICCV.2019.00630", "arxiv": "1812.03982" }, { "title": "Grounding DINO: Marrying DINO with Grounded Pre-Training for Open-Set Object Detection", "authors": "Shilong Liu, Zhaoyang Zeng, Tianhe Ren et al.", "venue": "ECCV", "year": 2023, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "In this paper, we present an open-set object detector, called Grounding DINO, by marrying Transformer-based detector DINO with grounded pre-training, which can detect arbitrary objects with human inputs such as category names or referring expressions. The key solution of open-set object detection is introducing language to a closed-set detector for open-set concept generalization. To effectively fuse language and vision modalities, we conceptually divide a closed-set detector into three phases and propose a tight fusion solution, which includes a feature enhancer, a language-guided query selection, and a cross-modality decoder for cross-modality fusion. While previous works mainly evaluate open-set object detection on novel categories, we propose to also perform evaluations on referring expression comprehension for objects specified with attributes. Grounding DINO performs remarkably well on all three settings, including benchmarks on COCO, LVIS, ODinW, and RefCOCO/+/g. Grounding DINO achieves a $52.5$ AP on the COCO detection zero-shot transfer benchmark, i.e., without any training data from COCO. It sets a new record on the ODinW zero-shot benchmark with a mean $26.1$ AP. Code will be available at \\url{https://github.com/IDEA-Research/GroundingDINO}.", "motivation": "In this paper, we present an open-set object detector, called Grounding DINO, by marrying Transformer-based detector DINO with grounded pre-training, which can detect arbitrary objects with human inputs such as category names or referring expressions.", "implementation": "While previous works mainly evaluate open-set object detection on novel categories, we propose to also perform evaluations on referring expression comprehension for objects specified with attributes.", "application": "The key solution of open-set object detection is introducing language to a closed-set detector for open-set concept generalization.", "problem": "In this paper, we present an open-set object detector, called Grounding DINO, by marrying Transformer-based detector DINO with grounded pre-training, which can detect arbitrary objects with human inputs such as category names or referring expressions.", "contribution": "While previous works mainly evaluate open-set object detection on novel categories, we propose to also perform evaluations on referring expression comprehension for objects specified with attributes.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Grounding DINO: Marrying DINO with Grounded Pre-Training for Open-Set Object Detection 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 4,062；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "ECCV", "Datasets / Benchmarks" ], "code": { "available": true, "url": "https://github.com/IDEA-Research/GroundingDINO}" }, "url": "http://arxiv.org/pdf/2303.05499", "updated": "2026-05-20", "topCitedRank": 66, "citationCount": 4062, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/c3e5a20b844c042d2174263d2fd5b30d8cc8f0b0", "doi": "10.48550/arXiv.2303.05499", "arxiv": "2303.05499" }, { "title": "Scalability in Perception for Autonomous Driving: Waymo Open Dataset", "authors": "Pei Sun, Henrik Kretzschmar, Xerxes Dotiwalla et al.", "venue": "CVPR", "year": 2019, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "The research community has increasing interest in autonomous driving research, despite the resource intensity of obtaining representative real world data. Existing self-driving datasets are limited in the scale and variation of the environments they capture, even though generalization within and between operating regions is crucial to the over-all viability of the technology. In an effort to help align the research community’s contributions with real-world self-driving problems, we introduce a new large scale, high quality, diverse dataset. Our new dataset consists of 1150 scenes that each span 20 seconds, consisting of well synchronized and calibrated high quality LiDAR and camera data captured across a range of urban and suburban geographies. It is 15x more diverse than the largest camera+LiDAR dataset available based on our proposed diversity metric. We exhaustively annotated this data with 2D (camera image) and 3D (LiDAR) bounding boxes, with consistent identifiers across frames. Finally, we provide strong baselines for 2D as well as 3D detection and tracking tasks. We further study the effects of dataset size and generalization across geographies on 3D detection methods. Find data, code and more up-to-date information at http://www.waymo.com/open.", "motivation": "The research community has increasing interest in autonomous driving research, despite the resource intensity of obtaining representative real world data.", "implementation": "In an effort to help align the research community’s contributions with real-world self-driving problems, we introduce a new large scale, high quality, diverse dataset.", "application": "Our new dataset consists of 1150 scenes that each span 20 seconds, consisting of well synchronized and calibrated high quality LiDAR and camera data captured across a range of urban and suburban geographies.", "problem": "The research community has increasing interest in autonomous driving research, despite the resource intensity of obtaining representative real world data.", "contribution": "In an effort to help align the research community’s contributions with real-world self-driving problems, we introduce a new large scale, high quality, diverse dataset.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Scalability in Perception for Autonomous Driving: Waymo Open Dataset 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 3,994；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Datasets / Benchmarks" ], "code": { "available": false, "url": "" }, "url": "https://arxiv.org/pdf/1912.04838", "updated": "2026-05-20", "topCitedRank": 67, "citationCount": 3994, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/8406903fd2f0eb25349bf071ccfaae3947e2a9cd", "doi": "10.1109/CVPR42600.2020.00252", "arxiv": "1912.04838" }, { "title": "A Discriminative Feature Learning Approach for Deep Face Recognition", "authors": "Yandong Wen, Kaipeng Zhang, Zhifeng Li et al.", "venue": "ECCV", "year": 2016, "topic": "Face & Biometrics", "modality": "Face imagery", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "人脸识别、表征学习或伪造检测", "method": "判别式嵌入损失、角度边界或检测网络", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "This paper improves deep face recognition by learning angularly discriminative features. It introduces an angular softmax loss that constrains features on a hypersphere and enforces a larger angular margin between face identities.", "motivation": "Face recognition needs features with small intra-class variation and large inter-class separation, while ordinary softmax classification does not explicitly impose a discriminative angular margin.", "implementation": "The method replaces the standard softmax objective with an angular-margin formulation on normalized features and weights, producing more separable face embeddings on the hypersphere.", "application": "Useful for face verification and identification benchmarks such as LFW, YTF, and MegaFace, and for biometric recognition systems requiring discriminative embeddings.", "problem": "Face recognition needs features with small intra-class variation and large inter-class separation, while ordinary softmax classification does not explicitly impose a discriminative angular margin.", "contribution": "The method replaces the standard softmax objective with an angular-margin formulation on normalized features and weights, producing more separable face embeddings on the hypersphere.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "A Discriminative Feature Learning Approach for Deep Face Recognition 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Face & Biometrics 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 3,981；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何同时处理隐私、公平性、跨域泛化和安全攻击。", "directionNote": "Useful for face verification and identification benchmarks such as LFW, YTF, and MegaFace, and for biometric recognition systems requiring discriminative embeddings.", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "ECCV", "Face & Biometrics" ], "code": { "available": false, "url": "" }, "url": "https://www.semanticscholar.org/paper/4cfd770ccecae1c0b4248bc800d7fd35c817bbbd", "updated": "2026-05-20", "topCitedRank": 68, "citationCount": 3981, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/4cfd770ccecae1c0b4248bc800d7fd35c817bbbd", "doi": "10.1007/978-3-319-46478-7_31", "arxiv": "" }, { "title": "GhostNet: More Features From Cheap Operations", "authors": "Kai Han, Yunhe Wang, Qi Tian et al.", "venue": "CVPR", "year": 2019, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "Deploying convolutional neural networks (CNNs) on embedded devices is difficult due to the limited memory and computation resources. The redundancy in feature maps is an important characteristic of those successful CNNs, but has rarely been investigated in neural architecture design. This paper proposes a novel Ghost module to generate more feature maps from cheap operations. Based on a set of intrinsic feature maps, we apply a series of linear transformations with cheap cost to generate many ghost feature maps that could fully reveal information underlying intrinsic features. The proposed Ghost module can be taken as a plug-and-play component to upgrade existing convolutional neural networks. Ghost bottlenecks are designed to stack Ghost modules, and then the lightweight GhostNet can be easily established. Experiments conducted on benchmarks demonstrate that the proposed Ghost module is an impressive alternative of convolution layers in baseline models, and our GhostNet can achieve higher recognition performance (e.g. 75.7% top-1 accuracy) than MobileNetV3 with similar computational cost on the ImageNet ILSVRC-2012 classification dataset. Code is available at https://github.com/huawei-noah/ghostnet.", "motivation": "Deploying convolutional neural networks (CNNs) on embedded devices is difficult due to the limited memory and computation resources.", "implementation": "The redundancy in feature maps is an important characteristic of those successful CNNs, but has rarely been investigated in neural architecture design.", "application": "Experiments conducted on benchmarks demonstrate that the proposed Ghost module is an impressive alternative of convolution layers in baseline models, and our GhostNet can achieve higher recognition performance (e.g.", "problem": "Deploying convolutional neural networks (CNNs) on embedded devices is difficult due to the limited memory and computation resources.", "contribution": "The redundancy in feature maps is an important characteristic of those successful CNNs, but has rarely been investigated in neural architecture design.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "GhostNet: More Features From Cheap Operations 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 3,959；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Datasets / Benchmarks" ], "code": { "available": true, "url": "https://github.com/huawei-noah/ghostnet" }, "url": "https://arxiv.org/pdf/1911.11907", "updated": "2026-05-20", "topCitedRank": 69, "citationCount": 3959, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/a4cc0701170331a1fd0e58bad962bd7f39f5efc9", "doi": "10.1109/cvpr42600.2020.00165", "arxiv": "1911.11907" }, { "title": "Masked-attention Mask Transformer for Universal Image Segmentation", "authors": "Bowen Cheng, Ishan Misra, A. Schwing et al.", "venue": "CVPR", "year": 2021, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "Image segmentation groups pixels with different semantics, e.g., category or instance membership. Each choice of semantics defines a task. While only the semantics of each task differ, current research focuses on designing spe-cialized architectures for each task. We present Masked- attention Mask Transformer (Mask2Former), a new archi-tecture capable of addressing any image segmentation task (panoptic, instance or semantic). Its key components in-clude masked attention, which extracts localized features by constraining cross-attention within predicted mask regions. In addition to reducing the research effort by at least three times, it outperforms the best specialized architectures by a significant margin on four popular datasets. Most no-tably, Mask2Former sets a new state-of-the-art for panoptic segmentation (57.8 PQ on COCO), instance segmentation (50.1 AP on COCO) and semantic segmentation (57.7 mIoU onADE20K).", "motivation": "Image segmentation groups pixels with different semantics, e.g., category or instance membership.", "implementation": "While only the semantics of each task differ, current research focuses on designing spe-cialized architectures for each task.", "application": "We present Masked- attention Mask Transformer (Mask2Former), a new archi-tecture capable of addressing any image segmentation task (panoptic, instance or semantic).", "problem": "Image segmentation groups pixels with different semantics, e.g., category or instance membership.", "contribution": "While only the semantics of each task differ, current research focuses on designing spe-cialized architectures for each task.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Masked-attention Mask Transformer for Universal Image Segmentation 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 3,863；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Datasets / Benchmarks" ], "code": { "available": false, "url": "" }, "url": "http://arxiv.org/pdf/2112.01527", "updated": "2026-05-20", "topCitedRank": 70, "citationCount": 3863, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/658a017302d29e4acf4ca789cb5d9f27983717ff", "doi": "10.1109/CVPR52688.2022.00135", "arxiv": "2112.01527" }, { "title": "Restormer: Efficient Transformer for High-Resolution Image Restoration", "authors": "Syed Waqas Zamir, Aditya Arora, Salman Hameed Khan et al.", "venue": "CVPR", "year": 2021, "topic": "Video / Human Understanding", "modality": "Video / human-centric data", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "视频理解、动作识别、人体姿态或轨迹预测", "method": "时序建模、人体结构先验或视频 transformer", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "Since convolutional neural networks (CNNs) perform well at learning generalizable image priors from large-scale data, these models have been extensively applied to image restoration and related tasks. Recently, another class of neural architectures, Transformers, have shown significant performance gains on natural language and high-level vision tasks. While the Transformer model mitigates the shortcomings of CNNs (i.e., limited receptive field and inadaptability to input content), its computational complexity grows quadratically with the spatial resolution, therefore making it infeasible to apply to most image restoration tasks involving high-resolution images. In this work, we propose an efficient Transformer model by making several key designs in the building blocks (multi-head attention and feed-forward network) such that it can capture long-range pixel interactions, while still remaining applicable to large images. Our model, named Restoration Transformer (Restormer), achieves state-of-the-art results on several image restoration tasks, including image deraining, single-image motion deblurring, defocus deblurring (single-image and dual-pixel data), and image denoising (Gaussian grayscale/color denoising, and real image denoising). The source code and pre-trained models are available at https://github.com/swz30/Restormer.", "motivation": "While the Transformer model mitigates the shortcomings of CNNs (i.e., limited receptive field and inadaptability to input content), its computational complexity grows quadratically with the spatial resolution, therefore making it infeasible to apply to most image restoration tasks involving high-resolution images.", "implementation": "In this work, we propose an efficient Transformer model by making several key designs in the building blocks (multi-head attention and feed-forward network) such that it can capture long-range pixel interactions, while still remaining applicable to large images.", "application": "Our model, named Restoration Transformer (Restormer), achieves state-of-the-art results on several image restoration tasks, including image deraining, single-image motion deblurring, defocus deblurring (single-image and dual-pixel data), and image denoising (Gaussian grayscale/color denoising, and real image denoising).", "problem": "While the Transformer model mitigates the shortcomings of CNNs (i.e., limited receptive field and inadaptability to input content), its computational complexity grows quadratically with the spatial resolution, therefore making it infeasible to apply to most image restoration tasks involving high-resolution images.", "contribution": "In this work, we propose an efficient Transformer model by making several key designs in the building blocks (multi-head attention and feed-forward network) such that it can capture long-range pixel interactions, while still remaining applicable to large images.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Restormer: Efficient Transformer for High-Resolution Image Restoration 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Video / Human Understanding 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 3,846；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何在长时序、遮挡和复杂交互下保持鲁棒理解。", "directionNote": "该方向的高引用工作集中体现了 Video / Human Understanding 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Video / Human Understanding" ], "code": { "available": true, "url": "https://github.com/swz30/Restormer" }, "url": "https://arxiv.org/pdf/2111.09881", "updated": "2026-05-20", "topCitedRank": 71, "citationCount": 3846, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/1e88d5afe19aea324d33541f60a90b7036894c32", "doi": "10.1109/CVPR52688.2022.00564", "arxiv": "2111.09881" }, { "title": "Scene Parsing through ADE20K Dataset", "authors": "Bolei Zhou, Hang Zhao, Xavier Puig et al.", "venue": "CVPR", "year": 2017, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "This paper introduces ADE20K for scene parsing: a densely annotated dataset covering diverse scenes, objects, and object parts. It uses the dataset to evaluate semantic segmentation and scene parsing models.", "motivation": "Scene parsing requires recognizing both objects and stuff across diverse indoor and outdoor scenes, but earlier datasets were limited in category coverage, density, or scene diversity.", "implementation": "The work builds ADE20K with dense pixel-level annotations for scenes, objects, and parts, then establishes benchmark evaluations for semantic segmentation and scene parsing.", "application": "Useful for semantic segmentation, scene understanding, object-and-part parsing, dataset benchmarking, and training perception models for complex real-world environments.", "problem": "Scene parsing requires recognizing both objects and stuff across diverse indoor and outdoor scenes, but earlier datasets were limited in category coverage, density, or scene diversity.", "contribution": "The work builds ADE20K with dense pixel-level annotations for scenes, objects, and parts, then establishes benchmark evaluations for semantic segmentation and scene parsing.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Scene Parsing through ADE20K Dataset 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 3,822；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "Useful for semantic segmentation, scene understanding, object-and-part parsing, dataset benchmarking, and training perception models for complex real-world environments.", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Datasets / Benchmarks" ], "code": { "available": false, "url": "" }, "url": "https://www.semanticscholar.org/paper/2a5667702b0f1ff77dde8fb3e2e10d4e05e8de9d", "updated": "2026-05-20", "topCitedRank": 72, "citationCount": 3822, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/2a5667702b0f1ff77dde8fb3e2e10d4e05e8de9d", "doi": "10.1109/CVPR.2017.544", "arxiv": "" }, { "title": "RAFT: Recurrent All-Pairs Field Transforms for Optical Flow", "authors": "Zachary Teed, Jia Deng", "venue": "ECCV", "year": 2020, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "High", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "We introduce Recurrent All-Pairs Field Transforms (RAFT), a new deep network architecture for optical flow. RAFT extracts per-pixel features, builds multi-scale 4D correlation volumes for all pairs of pixels, and iteratively updates a flow field through a recurrent unit that performs lookups on the correlation volumes. RAFT achieves state-of-the-art performance on the KITTI and Sintel datasets. In addition, RAFT has strong cross-dataset generalization as well as high efficiency in inference time, training speed, and parameter count.", "motivation": "We introduce Recurrent All-Pairs Field Transforms (RAFT), a new deep network architecture for optical flow.", "implementation": "In addition, RAFT has strong cross-dataset generalization as well as high efficiency in inference time, training speed, and parameter count.", "application": "RAFT achieves state-of-the-art performance on the KITTI and Sintel datasets.", "problem": "We introduce Recurrent All-Pairs Field Transforms (RAFT), a new deep network architecture for optical flow.", "contribution": "In addition, RAFT has strong cross-dataset generalization as well as high efficiency in inference time, training speed, and parameter count.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "RAFT: Recurrent All-Pairs Field Transforms for Optical Flow 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 3,729；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "ECCV", "Datasets / Benchmarks", "award-winning", "Best Paper" ], "code": { "available": false, "url": "" }, "url": "https://www.semanticscholar.org/paper/3230e2d6b4671cc03974af2219c6d3270e6fac70", "updated": "2026-05-20", "topCitedRank": 73, "citationCount": 3729, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/3230e2d6b4671cc03974af2219c6d3270e6fac70", "doi": "10.1007/978-3-030-58536-5_24", "arxiv": "2003.12039", "awards": [ { "conference": "ECCV", "year": 2020, "award": "Best Paper", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ] }, { "title": "Colorful Image Colorization", "authors": "Richard Zhang, Phillip Isola, Alexei A. Efros", "venue": "ECCV", "year": 2016, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "Given a grayscale photograph as input, this paper attacks the problem of hallucinating a plausible color version of the photograph. This problem is clearly underconstrained, so previous approaches have either relied on significant user interaction or resulted in desaturated colorizations. We propose a fully automatic approach that produces vibrant and realistic colorizations. We embrace the underlying uncertainty of the problem by posing it as a classification task and use class-rebalancing at training time to increase the diversity of colors in the result. The system is implemented as a feed-forward pass in a CNN at test time and is trained on over a million color images. We evaluate our algorithm using a “colorization Turing test,” asking human participants to choose between a generated and ground truth color image. Our method successfully fools humans on 32 % of the trials, significantly higher than previous methods. Moreover, we show that colorization can be a powerful pretext task for self-supervised feature learning, acting as a cross-channel encoder. This approach results in state-of-the-art performance on several feature learning benchmarks.", "motivation": "Given a grayscale photograph as input, this paper attacks the problem of hallucinating a plausible color version of the photograph.", "implementation": "We propose a fully automatic approach that produces vibrant and realistic colorizations.", "application": "We embrace the underlying uncertainty of the problem by posing it as a classification task and use class-rebalancing at training time to increase the diversity of colors in the result.", "problem": "Given a grayscale photograph as input, this paper attacks the problem of hallucinating a plausible color version of the photograph.", "contribution": "We propose a fully automatic approach that produces vibrant and realistic colorizations.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Colorful Image Colorization 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 3,718；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "ECCV", "Datasets / Benchmarks" ], "code": { "available": false, "url": "" }, "url": "https://link.springer.com/content/pdf/10.1007/978-3-319-46487-9_40.pdf", "updated": "2026-05-20", "topCitedRank": 74, "citationCount": 3718, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/8201e6e687f2de477258e9be53ba7b73ee30d7de", "doi": "10.1007/978-3-319-46487-9_40", "arxiv": "1603.08511" }, { "title": "Rethinking Semantic Segmentation from a Sequence-to-Sequence Perspective with Transformers", "authors": "Sixiao Zheng, Jiachen Lu, Hengshuang Zhao et al.", "venue": "CVPR", "year": 2020, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "Most recent semantic segmentation methods adopt a fully-convolutional network (FCN) with an encoder-decoder architecture. The encoder progressively reduces the spatial resolution and learns more abstract/semantic visual concepts with larger receptive fields. Since context modeling is critical for segmentation, the latest efforts have been focused on increasing the receptive field, through either dilated/atrous convolutions or inserting attention modules. However, the encoder-decoder based FCN architecture remains unchanged. In this paper, we aim to provide an alternative perspective by treating semantic segmentation as a sequence-to-sequence prediction task. Specifically, we deploy a pure transformer (i.e., without convolution and resolution reduction) to encode an image as a sequence of patches. With the global context modeled in every layer of the transformer, this encoder can be combined with a simple decoder to provide a powerful segmentation model, termed SEgmentation TRansformer (SETR). Extensive experiments show that SETR achieves new state of the art on ADE20K (50.28% mIoU), Pascal Context (55.83% mIoU) and competitive results on Cityscapes. Particularly, we achieve the first position in the highly competitive ADE20K test server leaderboard on the day of submission.", "motivation": "However, the encoder-decoder based FCN architecture remains unchanged.", "implementation": "Most recent semantic segmentation methods adopt a fully-convolutional network (FCN) with an encoder-decoder architecture.", "application": "Since context modeling is critical for segmentation, the latest efforts have been focused on increasing the receptive field, through either dilated/atrous convolutions or inserting attention modules.", "problem": "However, the encoder-decoder based FCN architecture remains unchanged.", "contribution": "Most recent semantic segmentation methods adopt a fully-convolutional network (FCN) with an encoder-decoder architecture.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Rethinking Semantic Segmentation from a Sequence-to-Sequence Perspective with Transformers 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 3,621；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Datasets / Benchmarks" ], "code": { "available": false, "url": "" }, "url": "https://arxiv.org/pdf/2012.15840", "updated": "2026-05-20", "topCitedRank": 75, "citationCount": 3621, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/d29430adccb805ab57b349afa8553954347b3197", "doi": "10.1109/CVPR46437.2021.00681", "arxiv": "2012.15840" }, { "title": "YOLOv9: Learning What You Want to Learn Using Programmable Gradient Information", "authors": "Chien-Yao Wang, I-Hau Yeh, Hongpeng Liao", "venue": "ECCV", "year": 2024, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "Today's deep learning methods focus on how to design the most appropriate objective functions so that the prediction results of the model can be closest to the ground truth. Meanwhile, an appropriate architecture that can facilitate acquisition of enough information for prediction has to be designed. Existing methods ignore a fact that when input data undergoes layer-by-layer feature extraction and spatial transformation, large amount of information will be lost. This paper will delve into the important issues of data loss when data is transmitted through deep networks, namely information bottleneck and reversible functions. We proposed the concept of programmable gradient information (PGI) to cope with the various changes required by deep networks to achieve multiple objectives. PGI can provide complete input information for the target task to calculate objective function, so that reliable gradient information can be obtained to update network weights. In addition, a new lightweight network architecture -- Generalized Efficient Layer Aggregation Network (GELAN), based on gradient path planning is designed. GELAN's architecture confirms that PGI has gained superior results on lightweight models. We verified the proposed GELAN and PGI on MS COCO dataset based object detection. The results show that GELAN only uses conventional convolution operators to achieve better parameter utilization than the state-of-the-art methods developed based on depth-wise convolution. PGI can be used for variety of models from lightweight to large. It can be used to obtain complete information, so that train-from-scratch models can achieve better results than state-of-the-art models pre-trained using large datasets, the comparison results are shown in Figure 1. The source codes are at: https://github.com/WongKinYiu/yolov9.", "motivation": "This paper will delve into the important issues of data loss when data is transmitted through deep networks, namely information bottleneck and reversible functions.", "implementation": "Today's deep learning methods focus on how to design the most appropriate objective functions so that the prediction results of the model can be closest to the ground truth.", "application": "We verified the proposed GELAN and PGI on MS COCO dataset based object detection.", "problem": "This paper will delve into the important issues of data loss when data is transmitted through deep networks, namely information bottleneck and reversible functions.", "contribution": "Today's deep learning methods focus on how to design the most appropriate objective functions so that the prediction results of the model can be closest to the ground truth.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "YOLOv9: Learning What You Want to Learn Using Programmable Gradient Information 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 3,620；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "ECCV", "Datasets / Benchmarks" ], "code": { "available": true, "url": "https://github.com/WongKinYiu/yolov9" }, "url": "https://www.semanticscholar.org/paper/cf70392a3b1ae92fdb1b70448aaddcbd03726d3d", "updated": "2026-05-20", "topCitedRank": 76, "citationCount": 3620, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/cf70392a3b1ae92fdb1b70448aaddcbd03726d3d", "doi": "10.48550/arXiv.2402.13616", "arxiv": "2402.13616" }, { "title": "Residual Attention Network for Image Classification", "authors": "Fei Wang, Mengqing Jiang, C. Qian et al.", "venue": "CVPR", "year": 2017, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "In this work, we propose Residual Attention Network, a convolutional neural network using attention mechanism which can incorporate with state-of-art feed forward network architecture in an end-to-end training fashion. Our Residual Attention Network is built by stacking Attention Modules which generate attention-aware features. The attention-aware features from different modules change adaptively as layers going deeper. Inside each Attention Module, bottom-up top-down feedforward structure is used to unfold the feedforward and feedback attention process into a single feedforward process. Importantly, we propose attention residual learning to train very deep Residual Attention Networks which can be easily scaled up to hundreds of layers. Extensive analyses are conducted on CIFAR-10 and CIFAR-100 datasets to verify the effectiveness of every module mentioned above. Our Residual Attention Network achieves state-of-the-art object recognition performance on three benchmark datasets including CIFAR-10 (3.90% error), CIFAR-100 (20.45% error) and ImageNet (4.8% single model and single crop, top-5 error). Note that, our method achieves 0.6% top-1 accuracy improvement with 46% trunk depth and 69% forward FLOPs comparing to ResNet-200. The experiment also demonstrates that our network is robust against noisy labels.", "motivation": "In this work, we propose Residual Attention Network, a convolutional neural network using attention mechanism which can incorporate with state-of-art feed forward network architecture in an end-to-end training fashion.", "implementation": "Our Residual Attention Network is built by stacking Attention Modules which generate attention-aware features.", "application": "Inside each Attention Module, bottom-up top-down feedforward structure is used to unfold the feedforward and feedback attention process into a single feedforward process.", "problem": "In this work, we propose Residual Attention Network, a convolutional neural network using attention mechanism which can incorporate with state-of-art feed forward network architecture in an end-to-end training fashion.", "contribution": "Our Residual Attention Network is built by stacking Attention Modules which generate attention-aware features.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Residual Attention Network for Image Classification 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 3,600；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Datasets / Benchmarks" ], "code": { "available": false, "url": "" }, "url": "https://arxiv.org/pdf/1704.06904", "updated": "2026-05-20", "topCitedRank": 77, "citationCount": 3600, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/77d30cf9a34fb6b50979c6a68863099da9a060ad", "doi": "10.1109/CVPR.2017.683", "arxiv": "1704.06904" }, { "title": "MnasNet: Platform-Aware Neural Architecture Search for Mobile", "authors": "Mingxing Tan, Bo Chen, Ruoming Pang et al.", "venue": "CVPR", "year": 2018, "topic": "Object Detection", "modality": "RGB / multi-scale visual features", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "目标检测、开放集检测或检测损失设计", "method": "检测框架、特征金字塔、transformer detector 或检测损失", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "Designing convolutional neural networks (CNN) for mobile devices is challenging because mobile models need to be small and fast, yet still accurate. Although significant efforts have been dedicated to design and improve mobile CNNs on all dimensions, it is very difficult to manually balance these trade-offs when there are so many architectural possibilities to consider. In this paper, we propose an automated mobile neural architecture search (MNAS) approach, which explicitly incorporate model latency into the main objective so that the search can identify a model that achieves a good trade-off between accuracy and latency. Unlike previous work, where latency is considered via another, often inaccurate proxy (e.g., FLOPS), our approach directly measures real-world inference latency by executing the model on mobile phones. To further strike the right balance between flexibility and search space size, we propose a novel factorized hierarchical search space that encourages layer diversity throughout the network. Experimental results show that our approach consistently outperforms state-of-the-art mobile CNN models across multiple vision tasks. On the ImageNet classification task, our MnasNet achieves 75.2% top-1 accuracy with 78ms latency on a Pixel phone, which is 1.8× faster than MobileNetV2 with 0.5% higher accuracy and 2.3× faster than NASNet with 1.2% higher accuracy. Our MnasNet also achieves better mAP quality than MobileNets for COCO object detection. Code is at https://github.com/tensorflow/tpu/tree/master/models/official/mnasnet.", "motivation": "Designing convolutional neural networks (CNN) for mobile devices is challenging because mobile models need to be small and fast, yet still accurate.", "implementation": "In this paper, we propose an automated mobile neural architecture search (MNAS) approach, which explicitly incorporate model latency into the main objective so that the search can identify a model that achieves a good trade-off between accuracy and latency.", "application": "Experimental results show that our approach consistently outperforms state-of-the-art mobile CNN models across multiple vision tasks.", "problem": "Designing convolutional neural networks (CNN) for mobile devices is challenging because mobile models need to be small and fast, yet still accurate.", "contribution": "In this paper, we propose an automated mobile neural architecture search (MNAS) approach, which explicitly incorporate model latency into the main objective so that the search can identify a model that achieves a good trade-off between accuracy and latency.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "MnasNet: Platform-Aware Neural Architecture Search for Mobile 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Object Detection 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 3,394；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何在开放世界、长尾类别和实时部署中保持稳定检测性能。", "directionNote": "该方向的高引用工作集中体现了 Object Detection 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Object Detection" ], "code": { "available": true, "url": "https://github.com/tensorflow/tpu/tree/master/models/official/mnasnet" }, "url": "https://arxiv.org/pdf/1807.11626", "updated": "2026-05-20", "topCitedRank": 78, "citationCount": 3394, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/693c97ecedb0a84539b7162c95e89fa3cd84ca73", "doi": "10.1109/CVPR.2019.00293", "arxiv": "1807.11626" }, { "title": "Accelerating the Super-Resolution Convolutional Neural Network", "authors": "Chao Dong, Chen Change Loy, Xiaoou Tang", "venue": "ECCV", "year": 2016, "topic": "3D / Geometry / Driving", "modality": "3D / depth / optical flow / driving sensors", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "三维理解、几何重建、光流、深度或自动驾驶感知", "method": "几何约束、点云网络、连续场表示或多传感器数据集", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "As a successful deep model applied in image super-resolution (SR), the Super-Resolution Convolutional Neural Network (SRCNN) [1, 2] has demonstrated superior performance to the previous hand-crafted models either in speed and restoration quality. However, the high computational cost still hinders it from practical usage that demands real-time performance (24 fps). In this paper, we aim at accelerating the current SRCNN, and propose a compact hourglass-shape CNN structure for faster and better SR. We re-design the SRCNN structure mainly in three aspects. First, we introduce a deconvolution layer at the end of the network, then the mapping is learned directly from the original low-resolution image (without interpolation) to the high-resolution one. Second, we reformulate the mapping layer by shrinking the input feature dimension before mapping and expanding back afterwards. Third, we adopt smaller filter sizes but more mapping layers. The proposed model achieves a speed up of more than 40 times with even superior restoration quality. Further, we present the parameter settings that can achieve real-time performance on a generic CPU while still maintaining good performance. A corresponding transfer strategy is also proposed for fast training and testing across different upscaling factors.", "motivation": "However, the high computational cost still hinders it from practical usage that demands real-time performance (24 fps).", "implementation": "As a successful deep model applied in image super-resolution (SR), the Super-Resolution Convolutional Neural Network (SRCNN) [1, 2] has demonstrated superior performance to the previous hand-crafted models either in speed and restoration quality.", "application": "First, we introduce a deconvolution layer at the end of the network, then the mapping is learned directly from the original low-resolution image (without interpolation) to the high-resolution one.", "problem": "However, the high computational cost still hinders it from practical usage that demands real-time performance (24 fps).", "contribution": "As a successful deep model applied in image super-resolution (SR), the Super-Resolution Convolutional Neural Network (SRCNN) [1, 2] has demonstrated superior performance to the previous hand-crafted models either in speed and restoration quality.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Accelerating the Super-Resolution Convolutional Neural Network 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 3D / Geometry / Driving 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 3,389；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何把真实传感噪声、动态场景和跨域泛化纳入可靠评测。", "directionNote": "该方向的高引用工作集中体现了 3D / Geometry / Driving 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "ECCV", "3D / Geometry / Driving" ], "code": { "available": false, "url": "" }, "url": "https://www.semanticscholar.org/paper/79da740db9006b2aa3e7b571d038ec895e323121", "updated": "2026-05-20", "topCitedRank": 79, "citationCount": 3389, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/79da740db9006b2aa3e7b571d038ec895e323121", "doi": "10.1007/978-3-319-46475-6_25", "arxiv": "1608.00367" }, { "title": "Social LSTM: Human Trajectory Prediction in Crowded Spaces", "authors": "Alexandre Alahi, Kratarth Goel, Vignesh Ramanathan et al.", "venue": "CVPR", "year": 2016, "topic": "Video / Human Understanding", "modality": "Video / human-centric data", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "视频理解、动作识别、人体姿态或轨迹预测", "method": "时序建模、人体结构先验或视频 transformer", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "This paper models human trajectory prediction in crowded spaces with recurrent networks. It introduces Social LSTM, where each person has an LSTM and nearby hidden states are pooled to capture social interactions.", "motivation": "Pedestrian motion in crowds is not independent: future trajectories depend on nearby people, collision avoidance, group behavior, and other social interactions that simple motion models miss.", "implementation": "The method assigns an LSTM to each tracked person and uses a social pooling layer over neighboring hidden states so trajectory prediction can condition on local crowd interactions.", "application": "Useful for pedestrian trajectory forecasting, autonomous driving, mobile robotics, surveillance, crowd analysis, and human-aware navigation in shared spaces.", "problem": "Pedestrian motion in crowds is not independent: future trajectories depend on nearby people, collision avoidance, group behavior, and other social interactions that simple motion models miss.", "contribution": "The method assigns an LSTM to each tracked person and uses a social pooling layer over neighboring hidden states so trajectory prediction can condition on local crowd interactions.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Social LSTM: Human Trajectory Prediction in Crowded Spaces 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Video / Human Understanding 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 3,370；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何在长时序、遮挡和复杂交互下保持鲁棒理解。", "directionNote": "Useful for pedestrian trajectory forecasting, autonomous driving, mobile robotics, surveillance, crowd analysis, and human-aware navigation in shared spaces.", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Video / Human Understanding" ], "code": { "available": false, "url": "" }, "url": "https://infoscience.epfl.ch/record/230265/files/CVPR16_N_LSTM.pdf", "updated": "2026-05-20", "topCitedRank": 80, "citationCount": 3370, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/e11a020f0d2942d09127daf1ce7e658d3bf67291", "doi": "10.1109/CVPR.2016.110", "arxiv": "" }, { "title": "CenterNet: Keypoint Triplets for Object Detection", "authors": "Kaiwen Duan, S. Bai, Lingxi Xie et al.", "venue": "ICCV", "year": 2019, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "In object detection, keypoint-based approaches often experience the drawback of a large number of incorrect object bounding boxes, arguably due to the lack of an additional assessment inside cropped regions. This paper presents an efficient solution that explores the visual patterns within individual cropped regions with minimal costs. We build our framework upon a representative one-stage keypoint-based detector named CornerNet. Our approach, named CenterNet, detects each object as a triplet, rather than a pair, of keypoints, which improves both precision and recall. Accordingly, we design two customized modules, cascade corner pooling, and center pooling, that enrich information collected by both the top-left and bottom-right corners and provide more recognizable information from the central regions. On the MS-COCO dataset, CenterNet achieves an AP of 47.0 %, outperforming all existing one-stage detectors by at least 4.9%. Furthermore, with a faster inference speed than the top-ranked two-stage detectors, CenterNet demonstrates a comparable performance to these detectors. Code is available at https://github.com/Duankaiwen/CenterNet.", "motivation": "In object detection, keypoint-based approaches often experience the drawback of a large number of incorrect object bounding boxes, arguably due to the lack of an additional assessment inside cropped regions.", "implementation": "We build our framework upon a representative one-stage keypoint-based detector named CornerNet.", "application": "On the MS-COCO dataset, CenterNet achieves an AP of 47.0 %, outperforming all existing one-stage detectors by at least 4.9%.", "problem": "In object detection, keypoint-based approaches often experience the drawback of a large number of incorrect object bounding boxes, arguably due to the lack of an additional assessment inside cropped regions.", "contribution": "We build our framework upon a representative one-stage keypoint-based detector named CornerNet.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "CenterNet: Keypoint Triplets for Object Detection 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 3,364；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "ICCV", "Datasets / Benchmarks" ], "code": { "available": true, "url": "https://github.com/Duankaiwen/CenterNet" }, "url": "https://arxiv.org/pdf/1904.08189", "updated": "2026-05-20", "topCitedRank": 81, "citationCount": 3364, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/9662b48d9b8a8f2118487a7f3be2d76283848627", "doi": "10.1109/ICCV.2019.00667", "arxiv": "1904.08189" }, { "title": "FlowNet 2.0: Evolution of Optical Flow Estimation with Deep Networks", "authors": "Eddy Ilg, N. Mayer, Tonmoy Saikia et al.", "venue": "CVPR", "year": 2016, "topic": "3D / Geometry / Driving", "modality": "3D / depth / optical flow / driving sensors", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "三维理解、几何重建、光流、深度或自动驾驶感知", "method": "几何约束、点云网络、连续场表示或多传感器数据集", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "The FlowNet demonstrated that optical flow estimation can be cast as a learning problem. However, the state of the art with regard to the quality of the flow has still been defined by traditional methods. Particularly on small displacements and real-world data, FlowNet cannot compete with variational methods. In this paper, we advance the concept of end-to-end learning of optical flow and make it work really well. The large improvements in quality and speed are caused by three major contributions: first, we focus on the training data and show that the schedule of presenting data during training is very important. Second, we develop a stacked architecture that includes warping of the second image with intermediate optical flow. Third, we elaborate on small displacements by introducing a subnetwork specializing on small motions. FlowNet 2.0 is only marginally slower than the original FlowNet but decreases the estimation error by more than 50%. It performs on par with state-of-the-art methods, while running at interactive frame rates. Moreover, we present faster variants that allow optical flow computation at up to 140fps with accuracy matching the original FlowNet.", "motivation": "The FlowNet demonstrated that optical flow estimation can be cast as a learning problem.", "implementation": "Second, we develop a stacked architecture that includes warping of the second image with intermediate optical flow.", "application": "FlowNet 2.0 is only marginally slower than the original FlowNet but decreases the estimation error by more than 50%.", "problem": "The FlowNet demonstrated that optical flow estimation can be cast as a learning problem.", "contribution": "Second, we develop a stacked architecture that includes warping of the second image with intermediate optical flow.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "FlowNet 2.0: Evolution of Optical Flow Estimation with Deep Networks 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 3D / Geometry / Driving 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 3,361；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何把真实传感噪声、动态场景和跨域泛化纳入可靠评测。", "directionNote": "该方向的高引用工作集中体现了 3D / Geometry / Driving 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "3D / Geometry / Driving" ], "code": { "available": false, "url": "" }, "url": "https://arxiv.org/pdf/1612.01925", "updated": "2026-05-20", "topCitedRank": 82, "citationCount": 3361, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/edd846e76cacfba5be37da99c006e3ccc9b861b0", "doi": "10.1109/CVPR.2017.179", "arxiv": "1612.01925" }, { "title": "Occupancy Networks: Learning 3D Reconstruction in Function Space", "authors": "L. Mescheder, Michael Oechsle, M. Niemeyer et al.", "venue": "CVPR", "year": 2018, "topic": "3D / Geometry / Driving", "modality": "3D / depth / optical flow / driving sensors", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "三维理解、几何重建、光流、深度或自动驾驶感知", "method": "几何约束、点云网络、连续场表示或多传感器数据集", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "With the advent of deep neural networks, learning-based approaches for 3D reconstruction have gained popularity. However, unlike for images, in 3D there is no canonical representation which is both computationally and memory efficient yet allows for representing high-resolution geometry of arbitrary topology. Many of the state-of-the-art learning-based 3D reconstruction approaches can hence only represent very coarse 3D geometry or are limited to a restricted domain. In this paper, we propose Occupancy Networks, a new representation for learning-based 3D reconstruction methods. Occupancy networks implicitly represent the 3D surface as the continuous decision boundary of a deep neural network classifier. In contrast to existing approaches, our representation encodes a description of the 3D output at infinite resolution without excessive memory footprint. We validate that our representation can efficiently encode 3D structure and can be inferred from various kinds of input. Our experiments demonstrate competitive results, both qualitatively and quantitatively, for the challenging tasks of 3D reconstruction from single images, noisy point clouds and coarse discrete voxel grids. We believe that occupancy networks will become a useful tool in a wide variety of learning-based 3D tasks.", "motivation": "However, unlike for images, in 3D there is no canonical representation which is both computationally and memory efficient yet allows for representing high-resolution geometry of arbitrary topology.", "implementation": "In this paper, we propose Occupancy Networks, a new representation for learning-based 3D reconstruction methods.", "application": "With the advent of deep neural networks, learning-based approaches for 3D reconstruction have gained popularity.", "problem": "However, unlike for images, in 3D there is no canonical representation which is both computationally and memory efficient yet allows for representing high-resolution geometry of arbitrary topology.", "contribution": "In this paper, we propose Occupancy Networks, a new representation for learning-based 3D reconstruction methods.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Occupancy Networks: Learning 3D Reconstruction in Function Space 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 3D / Geometry / Driving 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 3,359；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何把真实传感噪声、动态场景和跨域泛化纳入可靠评测。", "directionNote": "该方向的高引用工作集中体现了 3D / Geometry / Driving 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "3D / Geometry / Driving" ], "code": { "available": false, "url": "" }, "url": "https://arxiv.org/pdf/1812.03828", "updated": "2026-05-20", "topCitedRank": 83, "citationCount": 3359, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/2e689bdce24cf3644432505ce2783f03a1445ed2", "doi": "10.1109/CVPR.2019.00459", "arxiv": "1812.03828" }, { "title": "ChestX-Ray8: Hospital-Scale Chest X-Ray Database and Benchmarks on Weakly-Supervised Classification and Localization of Common Thorax Diseases", "authors": "Xiaosong Wang, Yifan Peng, Le Lu et al.", "venue": "CVPR", "year": 2017, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "The chest X-ray is one of the most commonly accessible radiological examinations for screening and diagnosis of many lung diseases. A tremendous number of X-ray imaging studies accompanied by radiological reports are accumulated and stored in many modern hospitals Picture Archiving and Communication Systems (PACS). On the other side, it is still an open question how this type of hospital-size knowledge database containing invaluable imaging informatics (i.e., loosely labeled) can be used to facilitate the data-hungry deep learning paradigms in building truly large-scale high precision computer-aided diagnosis (CAD) systems. In this paper, we present a new chest X-ray database, namely ChestX-ray8, which comprises 108,948 frontal-view X-ray images of 32,717 unique patients with the text-mined eight disease image labels (where each image can have multi-labels), from the associated radiological reports using natural language processing. Importantly, we demonstrate that these commonly occurring thoracic diseases can be detected and even spatially-located via a unified weakly-supervised multi-label image classification and disease localization framework, which is validated using our proposed dataset. Although the initial quantitative results are promising as reported, deep convolutional neural network based reading chest X-rays (i.e., recognizing and locating the common disease patterns trained with only image-level labels) remains a strenuous task for fully-automated high precision CAD systems.", "motivation": "Although the initial quantitative results are promising as reported, deep convolutional neural network based reading chest X-rays (i.e., recognizing and locating the common disease patterns trained with only image-level labels) remains a strenuous task for fully-automated high precision CAD systems.", "implementation": "Importantly, we demonstrate that these commonly occurring thoracic diseases can be detected and even spatially-located via a unified weakly-supervised multi-label image classification and disease localization framework, which is validated using our proposed dataset.", "application": "On the other side, it is still an open question how this type of hospital-size knowledge database containing invaluable imaging informatics (i.e., loosely labeled) can be used to facilitate the data-hungry deep learning paradigms in building truly large-scale high precision computer-aided diagnosis (CAD) systems.", "problem": "Although the initial quantitative results are promising as reported, deep convolutional neural network based reading chest X-rays (i.e., recognizing and locating the common disease patterns trained with only image-level labels) remains a strenuous task for fully-automated high precision CAD systems.", "contribution": "Importantly, we demonstrate that these commonly occurring thoracic diseases can be detected and even spatially-located via a unified weakly-supervised multi-label image classification and disease localization framework, which is validated using our proposed dataset.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "ChestX-Ray8: Hospital-Scale Chest X-Ray Database and Benchmarks on Weakly-Supervised Classification and Localization of Common Thorax Diseases 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 3,353；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Datasets / Benchmarks" ], "code": { "available": false, "url": "" }, "url": "https://arxiv.org/pdf/1705.02315", "updated": "2026-05-20", "topCitedRank": 84, "citationCount": 3353, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/58b6bd06ea58c367c64286126ba14128b45041b8", "doi": "10.1109/CVPR.2017.369", "arxiv": "1705.02315" }, { "title": "Unsupervised Learning of Visual Representations by Solving Jigsaw Puzzles", "authors": "M. Noroozi, P. Favaro", "venue": "ECCV", "year": 2016, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "We propose a novel unsupervised learning approach to build features suitable for object detection and classification. The features are pre-trained on a large dataset without human annotation and later transferred via fine-tuning on a different, smaller and labeled dataset. The pre-training consists of solving jigsaw puzzles of natural images. To facilitate the transfer of features to other tasks, we introduce the context-free network (CFN), a siamese-ennead convolutional neural network. The features correspond to the columns of the CFN and they process image tiles independently (i.e., free of context). The later layers of the CFN then use the features to identify their geometric arrangement. Our experimental evaluations show that the learned features capture semantically relevant content. We pre-train the CFN on the training set of the ILSVRC2012 dataset and transfer the features on the combined training and validation set of Pascal VOC 2007 for object detection (via fast RCNN) and classification. These features outperform all current unsupervised features with \\(51.8\\,\\%\\) for detection and \\(68.6\\,\\%\\) for classification, and reduce the gap with supervised learning (\\(56.5\\,\\%\\) and \\(78.2\\,\\%\\) respectively).", "motivation": "We propose a novel unsupervised learning approach to build features suitable for object detection and classification.", "implementation": "The features are pre-trained on a large dataset without human annotation and later transferred via fine-tuning on a different, smaller and labeled dataset.", "application": "We pre-train the CFN on the training set of the ILSVRC2012 dataset and transfer the features on the combined training and validation set of Pascal VOC 2007 for object detection (via fast RCNN) and classification.", "problem": "We propose a novel unsupervised learning approach to build features suitable for object detection and classification.", "contribution": "The features are pre-trained on a large dataset without human annotation and later transferred via fine-tuning on a different, smaller and labeled dataset.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Unsupervised Learning of Visual Representations by Solving Jigsaw Puzzles 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 3,239；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "ECCV", "Datasets / Benchmarks" ], "code": { "available": false, "url": "" }, "url": "https://www.semanticscholar.org/paper/2ec8f7e0257a07d3914322b36072d1bbcd58a1e0", "updated": "2026-05-20", "topCitedRank": 85, "citationCount": 3239, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/2ec8f7e0257a07d3914322b36072d1bbcd58a1e0", "doi": "10.1007/978-3-319-46466-4_5", "arxiv": "1603.09246" }, { "title": "DETRs Beat YOLOs on Real-time Object Detection", "authors": "Wenyu Lv, Shangliang Xu, Yian Zhao et al.", "venue": "CVPR", "year": 2023, "topic": "Object Detection", "modality": "RGB / multi-scale visual features", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "目标检测、开放集检测或检测损失设计", "method": "检测框架、特征金字塔、transformer detector 或检测损失", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "The YOLO series has become the most popular frame-work for real-time object detection due to its reasonable trade-off between speed and accuracy. However, we observe that the speed and accuracy of YOLOs are negatively affected by the NMS. Recently, end-to-end Transformer-based detectors (DETRs) have provided an alternative to eliminating NMS. Nevertheless, the high computational cost limits their practicality and hinders them from fully exploiting the advantage of excluding NMS. In this paper, we propose the Real-Time DEtection TRansformer (RT-DETR), the first real-time end-to-end object detector to our best knowledge that addresses the above dilemma. We build RT-DETR in two steps, drawing on the advanced DETR: first we focus on maintaining accuracy while improving speed, followed by maintaining speed while improving accuracy. Specifically, we design an efficient hybrid encoder to expeditiously process multi-scale features by decoupling intra-scale interaction and cross-scale fusion to improve speed. Then, we propose the uncertainty-minimal query selection to provide high-quality initial queries to the decoder, thereby improving accuracy. In addition, RT-DETR supports flexible speed tuning by adjusting the number of decoder layers to adapt to various scenarios without retraining. Our RT-DETR-R50 /R101 achieves 53.1% 154.3% AP on COCO and 108 /74 FPS on T4 GPU, outperforming previously advanced YOLOs in both speed and accuracy. Furthermore, RT-DETR-R50 outperforms DINO-R50 by 2.2% AP in accuracy and about 21 times in FPS. After pre-training with Objects365, RT-DETR-R50 / R101 achieves 55.3% 156.2% AP. The project page: https://zhao-yian.github.io/RTDEtr.", "motivation": "However, we observe that the speed and accuracy of YOLOs are negatively affected by the NMS.", "implementation": "In this paper, we propose the Real-Time DEtection TRansformer (RT-DETR), the first real-time end-to-end object detector to our best knowledge that addresses the above dilemma.", "application": "The YOLO series has become the most popular frame-work for real-time object detection due to its reasonable trade-off between speed and accuracy.", "problem": "However, we observe that the speed and accuracy of YOLOs are negatively affected by the NMS.", "contribution": "In this paper, we propose the Real-Time DEtection TRansformer (RT-DETR), the first real-time end-to-end object detector to our best knowledge that addresses the above dilemma.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "DETRs Beat YOLOs on Real-time Object Detection 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Object Detection 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 3,224；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何在开放世界、长尾类别和实时部署中保持稳定检测性能。", "directionNote": "该方向的高引用工作集中体现了 Object Detection 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Object Detection" ], "code": { "available": false, "url": "" }, "url": "http://arxiv.org/pdf/2304.08069", "updated": "2026-05-20", "topCitedRank": 86, "citationCount": 3224, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/3a6c423cc0f687fc43a671e55dce06f681df7476", "doi": "10.1109/CVPR52733.2024.01605", "arxiv": "2304.08069" }, { "title": "KPConv: Flexible and Deformable Convolution for Point Clouds", "authors": "Hugues Thomas, C. Qi, Jean-Emmanuel Deschaud et al.", "venue": "ICCV", "year": 2019, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "We present Kernel Point Convolution (KPConv), a new design of point convolution, i.e. that operates on point clouds without any intermediate representation. The convolution weights of KPConv are located in Euclidean space by kernel points, and applied to the input points close to them. Its capacity to use any number of kernel points gives KPConv more flexibility than fixed grid convolutions. Furthermore, these locations are continuous in space and can be learned by the network. Therefore, KPConv can be extended to deformable convolutions that learn to adapt kernel points to local geometry. Thanks to a regular subsampling strategy, KPConv is also efficient and robust to varying densities. Whether they use deformable KPConv for complex tasks, or rigid KPconv for simpler tasks, our networks outperform state-of-the-art classification and segmentation approaches on several datasets. We also offer ablation studies and visualizations to provide understanding of what has been learned by KPConv and to validate the descriptive power of deformable KPConv.", "motivation": "We present Kernel Point Convolution (KPConv), a new design of point convolution, i.e.", "implementation": "The convolution weights of KPConv are located in Euclidean space by kernel points, and applied to the input points close to them.", "application": "Whether they use deformable KPConv for complex tasks, or rigid KPconv for simpler tasks, our networks outperform state-of-the-art classification and segmentation approaches on several datasets.", "problem": "We present Kernel Point Convolution (KPConv), a new design of point convolution, i.e.", "contribution": "The convolution weights of KPConv are located in Euclidean space by kernel points, and applied to the input points close to them.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "KPConv: Flexible and Deformable Convolution for Point Clouds 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 3,221；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "ICCV", "Datasets / Benchmarks" ], "code": { "available": false, "url": "" }, "url": "https://arxiv.org/pdf/1904.08889", "updated": "2026-05-20", "topCitedRank": 87, "citationCount": 3221, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/9218b3e7e447048a070e565b9add26e41d861964", "doi": "10.1109/ICCV.2019.00651", "arxiv": "1904.08889" }, { "title": "Unsupervised Monocular Depth Estimation with Left-Right Consistency", "authors": "Clément Godard, Oisin Mac Aodha, G. Brostow", "venue": "CVPR", "year": 2016, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "Learning based methods have shown very promising results for the task of depth estimation in single images. However, most existing approaches treat depth prediction as a supervised regression problem and as a result, require vast quantities of corresponding ground truth depth data for training. Just recording quality depth data in a range of environments is a challenging problem. In this paper, we innovate beyond existing approaches, replacing the use of explicit depth data during training with easier-to-obtain binocular stereo footage. We propose a novel training objective that enables our convolutional neural network to learn to perform single image depth estimation, despite the absence of ground truth depth data. Ex-ploiting epipolar geometry constraints, we generate disparity images by training our network with an image reconstruction loss. We show that solving for image reconstruction alone results in poor quality depth images. To overcome this problem, we propose a novel training loss that enforces consistency between the disparities produced relative to both the left and right images, leading to improved performance and robustness compared to existing approaches. Our method produces state of the art results for monocular depth estimation on the KITTI driving dataset, even outperforming supervised methods that have been trained with ground truth depth.", "motivation": "However, most existing approaches treat depth prediction as a supervised regression problem and as a result, require vast quantities of corresponding ground truth depth data for training.", "implementation": "We propose a novel training objective that enables our convolutional neural network to learn to perform single image depth estimation, despite the absence of ground truth depth data.", "application": "Learning based methods have shown very promising results for the task of depth estimation in single images.", "problem": "However, most existing approaches treat depth prediction as a supervised regression problem and as a result, require vast quantities of corresponding ground truth depth data for training.", "contribution": "We propose a novel training objective that enables our convolutional neural network to learn to perform single image depth estimation, despite the absence of ground truth depth data.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Unsupervised Monocular Depth Estimation with Left-Right Consistency 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 3,206；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Datasets / Benchmarks" ], "code": { "available": false, "url": "" }, "url": "https://arxiv.org/pdf/1609.03677", "updated": "2026-05-20", "topCitedRank": 88, "citationCount": 3206, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/4463dc4a32b948f0230f3b782cbfecaf1c9e5b1d", "doi": "10.1109/CVPR.2017.699", "arxiv": "1609.03677" }, { "title": "Multi-view 3D Object Detection Network for Autonomous Driving", "authors": "Xiaozhi Chen, Huimin Ma, Ji Wan et al.", "venue": "CVPR", "year": 2016, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "This paper aims at high-accuracy 3D object detection in autonomous driving scenario. We propose Multi-View 3D networks (MV3D), a sensory-fusion framework that takes both LIDAR point cloud and RGB images as input and predicts oriented 3D bounding boxes. We encode the sparse 3D point cloud with a compact multi-view representation. The network is composed of two subnetworks: one for 3D object proposal generation and another for multi-view feature fusion. The proposal network generates 3D candidate boxes efficiently from the birds eye view representation of 3D point cloud. We design a deep fusion scheme to combine region-wise features from multiple views and enable interactions between intermediate layers of different paths. Experiments on the challenging KITTI benchmark show that our approach outperforms the state-of-the-art by around 25% and 30% AP on the tasks of 3D localization and 3D detection. In addition, for 2D detection, our approach obtains 14.9% higher AP than the state-of-the-art on the hard data among the LIDAR-based methods.", "motivation": "Experiments on the challenging KITTI benchmark show that our approach outperforms the state-of-the-art by around 25% and 30% AP on the tasks of 3D localization and 3D detection.", "implementation": "We propose Multi-View 3D networks (MV3D), a sensory-fusion framework that takes both LIDAR point cloud and RGB images as input and predicts oriented 3D bounding boxes.", "application": "This paper aims at high-accuracy 3D object detection in autonomous driving scenario.", "problem": "Experiments on the challenging KITTI benchmark show that our approach outperforms the state-of-the-art by around 25% and 30% AP on the tasks of 3D localization and 3D detection.", "contribution": "We propose Multi-View 3D networks (MV3D), a sensory-fusion framework that takes both LIDAR point cloud and RGB images as input and predicts oriented 3D bounding boxes.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Multi-view 3D Object Detection Network for Autonomous Driving 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 3,147；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Datasets / Benchmarks" ], "code": { "available": false, "url": "" }, "url": "https://arxiv.org/pdf/1611.07759", "updated": "2026-05-20", "topCitedRank": 89, "citationCount": 3147, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/dc200ab22bf63e10e8b2af328a9e072d82cf75b7", "doi": "10.1109/CVPR.2017.691", "arxiv": "1611.07759" }, { "title": "Semantic Image Synthesis With Spatially-Adaptive Normalization", "authors": "Taesung Park, Ming-Yu Liu, Ting-Chun Wang et al.", "venue": "CVPR", "year": 2019, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "We propose spatially-adaptive normalization, a simple but effective layer for synthesizing photorealistic images given an input semantic layout. Previous methods directly feed the semantic layout as input to the network, forcing the network to memorize the information throughout all the layers. Instead, we propose using the input layout for modulating the activations in normalization layers through a spatially-adaptive, learned affine transformation. Experiments on several challenging datasets demonstrate the superiority of our method compared to existing approaches, regarding both visual fidelity and alignment with input layouts. Finally, our model allows users to easily control the style and content of image synthesis results as well as create multi-modal results. Code is available upon publication.", "motivation": "Experiments on several challenging datasets demonstrate the superiority of our method compared to existing approaches, regarding both visual fidelity and alignment with input layouts.", "implementation": "We propose spatially-adaptive normalization, a simple but effective layer for synthesizing photorealistic images given an input semantic layout.", "application": "Finally, our model allows users to easily control the style and content of image synthesis results as well as create multi-modal results.", "problem": "Experiments on several challenging datasets demonstrate the superiority of our method compared to existing approaches, regarding both visual fidelity and alignment with input layouts.", "contribution": "We propose spatially-adaptive normalization, a simple but effective layer for synthesizing photorealistic images given an input semantic layout.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Semantic Image Synthesis With Spatially-Adaptive Normalization 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 3,110；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Datasets / Benchmarks" ], "code": { "available": false, "url": "" }, "url": "https://arxiv.org/pdf/1903.07291", "updated": "2026-05-20", "topCitedRank": 90, "citationCount": 3110, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/a1a19aaddf57c0546357d890d9269092ba0afb26", "doi": "10.1109/CVPR.2019.00244", "arxiv": "1903.07291" }, { "title": "RefineNet: Multi-path Refinement Networks for High-Resolution Semantic Segmentation", "authors": "Guosheng Lin, Anton Milan, Chunhua Shen et al.", "venue": "CVPR", "year": 2016, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "Recently, very deep convolutional neural networks (CNNs) have shown outstanding performance in object recognition and have also been the first choice for dense classification problems such as semantic segmentation. However, repeated subsampling operations like pooling or convolution striding in deep CNNs lead to a significant decrease in the initial image resolution. Here, we present RefineNet, a generic multi-path refinement network that explicitly exploits all the information available along the down-sampling process to enable high-resolution prediction using long-range residual connections. In this way, the deeper layers that capture high-level semantic features can be directly refined using fine-grained features from earlier convolutions. The individual components of RefineNet employ residual connections following the identity mapping mindset, which allows for effective end-to-end training. Further, we introduce chained residual pooling, which captures rich background context in an efficient manner. We carry out comprehensive experiments and set new state-of-the-art results on seven public datasets. In particular, we achieve an intersection-over-union score of 83.4 on the challenging PASCAL VOC 2012 dataset, which is the best reported result to date.", "motivation": "However, repeated subsampling operations like pooling or convolution striding in deep CNNs lead to a significant decrease in the initial image resolution.", "implementation": "Here, we present RefineNet, a generic multi-path refinement network that explicitly exploits all the information available along the down-sampling process to enable high-resolution prediction using long-range residual connections.", "application": "Recently, very deep convolutional neural networks (CNNs) have shown outstanding performance in object recognition and have also been the first choice for dense classification problems such as semantic segmentation.", "problem": "However, repeated subsampling operations like pooling or convolution striding in deep CNNs lead to a significant decrease in the initial image resolution.", "contribution": "Here, we present RefineNet, a generic multi-path refinement network that explicitly exploits all the information available along the down-sampling process to enable high-resolution prediction using long-range residual connections.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "RefineNet: Multi-path Refinement Networks for High-Resolution Semantic Segmentation 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 3,091；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Datasets / Benchmarks" ], "code": { "available": false, "url": "" }, "url": "https://arxiv.org/pdf/1611.06612", "updated": "2026-05-20", "topCitedRank": 91, "citationCount": 3091, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/de4ee92cfad3734ca820d004bc9ee75fc9dcfbf4", "doi": "10.1109/CVPR.2017.549", "arxiv": "1611.06612" }, { "title": "SphereFace: Deep Hypersphere Embedding for Face Recognition", "authors": "Weiyang Liu, Yandong Wen, Zhiding Yu et al.", "venue": "CVPR", "year": 2017, "topic": "Video / Human Understanding", "modality": "Video / human-centric data", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "视频理解、动作识别、人体姿态或轨迹预测", "method": "时序建模、人体结构先验或视频 transformer", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "This paper addresses deep face recognition (FR) problem under open-set protocol, where ideal face features are expected to have smaller maximal intra-class distance than minimal inter-class distance under a suitably chosen metric space. However, few existing algorithms can effectively achieve this criterion. To this end, we propose the angular softmax (A-Softmax) loss that enables convolutional neural networks (CNNs) to learn angularly discriminative features. Geometrically, A-Softmax loss can be viewed as imposing discriminative constraints on a hypersphere manifold, which intrinsically matches the prior that faces also lie on a manifold. Moreover, the size of angular margin can be quantitatively adjusted by a parameter m. We further derive specific m to approximate the ideal feature criterion. Extensive analysis and experiments on Labeled Face in the Wild (LFW), Youtube Faces (YTF) and MegaFace Challenge 1 show the superiority of A-Softmax loss in FR tasks.", "motivation": "This paper addresses deep face recognition (FR) problem under open-set protocol, where ideal face features are expected to have smaller maximal intra-class distance than minimal inter-class distance under a suitably chosen metric space.", "implementation": "To this end, we propose the angular softmax (A-Softmax) loss that enables convolutional neural networks (CNNs) to learn angularly discriminative features.", "application": "Extensive analysis and experiments on Labeled Face in the Wild (LFW), Youtube Faces (YTF) and MegaFace Challenge 1 show the superiority of A-Softmax loss in FR tasks.", "problem": "This paper addresses deep face recognition (FR) problem under open-set protocol, where ideal face features are expected to have smaller maximal intra-class distance than minimal inter-class distance under a suitably chosen metric space.", "contribution": "To this end, we propose the angular softmax (A-Softmax) loss that enables convolutional neural networks (CNNs) to learn angularly discriminative features.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "SphereFace: Deep Hypersphere Embedding for Face Recognition 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Video / Human Understanding 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 3,053；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何在长时序、遮挡和复杂交互下保持鲁棒理解。", "directionNote": "该方向的高引用工作集中体现了 Video / Human Understanding 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Video / Human Understanding" ], "code": { "available": false, "url": "" }, "url": "http://arxiv.org/pdf/1704.08063", "updated": "2026-05-20", "topCitedRank": 92, "citationCount": 3053, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/bd8f77b7d3b9d272f7a68defc1412f73e5ac3135", "doi": "10.1109/CVPR.2017.713", "arxiv": "1704.08063" }, { "title": "GQA: A New Dataset for Real-World Visual Reasoning and Compositional Question Answering", "authors": "Drew A. Hudson, Christopher D. Manning", "venue": "CVPR", "year": 2019, "topic": "Vision-Language / Foundation Models", "modality": "Image-text / foundation model", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "视觉-语言理解、开放词表识别或基础模型适配", "method": "大规模预训练、多模态对齐或提示式视觉建模", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "We introduce GQA, a new dataset for real-world visual reasoning and compositional question answering, seeking to address key shortcomings of previous VQA datasets. We have developed a strong and robust question engine that leverages Visual Genome scene graph structures to create 22M diverse reasoning questions, which all come with functional programs that represent their semantics. We use the programs to gain tight control over the answer distribution and present a new tunable smoothing technique to mitigate question biases. Accompanying the dataset is a suite of new metrics that evaluate essential qualities such as consistency, grounding and plausibility. A careful analysis is performed for baselines as well as state-of-the-art models, providing fine-grained results for different question types and topologies. Whereas a blind LSTM obtains a mere 42.1%, and strong VQA models achieve 54.1%, human performance tops at 89.3%, offering ample opportunity for new research to explore. We hope GQA will provide an enabling resource for the next generation of models with enhanced robustness, improved consistency, and deeper semantic understanding of vision and language.", "motivation": "We introduce GQA, a new dataset for real-world visual reasoning and compositional question answering, seeking to address key shortcomings of previous VQA datasets.", "implementation": "Accompanying the dataset is a suite of new metrics that evaluate essential qualities such as consistency, grounding and plausibility.", "application": "We hope GQA will provide an enabling resource for the next generation of models with enhanced robustness, improved consistency, and deeper semantic understanding of vision and language.", "problem": "We introduce GQA, a new dataset for real-world visual reasoning and compositional question answering, seeking to address key shortcomings of previous VQA datasets.", "contribution": "Accompanying the dataset is a suite of new metrics that evaluate essential qualities such as consistency, grounding and plausibility.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "GQA: A New Dataset for Real-World Visual Reasoning and Compositional Question Answering 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Vision-Language / Foundation Models 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 3,053；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何降低标注和算力成本，同时保持可控性、可解释性和领域泛化。", "directionNote": "该方向的高引用工作集中体现了 Vision-Language / Foundation Models 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Vision-Language / Foundation Models" ], "code": { "available": false, "url": "" }, "url": "https://arxiv.org/pdf/1902.09506", "updated": "2026-05-20", "topCitedRank": 93, "citationCount": 3053, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/a7ac99d7cf3f568ab1a741392144b646b856ae0c", "doi": "10.1109/CVPR.2019.00686", "arxiv": "" }, { "title": "Class-Balanced Loss Based on Effective Number of Samples", "authors": "Yin Cui, Menglin Jia, Tsung-Yi Lin et al.", "venue": "CVPR", "year": 2019, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "With the rapid increase of large-scale, real-world datasets, it becomes critical to address the problem of long-tailed data distribution (i.e., a few classes account for most of the data, while most classes are under-represented). Existing solutions typically adopt class re-balancing strategies such as re-sampling and re-weighting based on the number of observations for each class. In this work, we argue that as the number of samples increases, the additional benefit of a newly added data point will diminish. We introduce a novel theoretical framework to measure data overlap by associating with each sample a small neighboring region rather than a single point. The effective number of samples is defined as the volume of samples and can be calculated by a simple formula $(1-\\beta^{n})/(1-\\beta)$, where $n$ is the number of samples and $\\beta \\in [0,1)$ is a hyperparameter. We design a re-weighting scheme that uses the effective number of samples for each class to re-balance the loss, thereby yielding a class-balanced loss. Comprehensive experiments are conducted on artificially induced long-tailed CIFAR datasets and large-scale datasets including ImageNet and iNaturalist. Our results show that when trained with the proposed class-balanced loss, the network is able to achieve significant performance gains on long-tailed datasets.", "motivation": "With the rapid increase of large-scale, real-world datasets, it becomes critical to address the problem of long-tailed data distribution (i.e., a few classes account for most of the data, while most classes are under-represented).", "implementation": "We introduce a novel theoretical framework to measure data overlap by associating with each sample a small neighboring region rather than a single point.", "application": "Comprehensive experiments are conducted on artificially induced long-tailed CIFAR datasets and large-scale datasets including ImageNet and iNaturalist.", "problem": "With the rapid increase of large-scale, real-world datasets, it becomes critical to address the problem of long-tailed data distribution (i.e., a few classes account for most of the data, while most classes are under-represented).", "contribution": "We introduce a novel theoretical framework to measure data overlap by associating with each sample a small neighboring region rather than a single point.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Class-Balanced Loss Based on Effective Number of Samples 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 3,004；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Datasets / Benchmarks" ], "code": { "available": false, "url": "" }, "url": "https://arxiv.org/pdf/1901.05555", "updated": "2026-05-20", "topCitedRank": 94, "citationCount": 3004, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/54036f43acc6c9b49b334270c7237217685f52fb", "doi": "10.1109/CVPR.2019.00949", "arxiv": "1901.05555" }, { "title": "Sigmoid Loss for Language Image Pre-Training", "authors": "Xiaohua Zhai, Basil Mustafa, Alexander Kolesnikov et al.", "venue": "ICCV", "year": 2023, "topic": "Vision-Language / Foundation Models", "modality": "Image-text / foundation model", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "视觉-语言理解、开放词表识别或基础模型适配", "method": "大规模预训练、多模态对齐或提示式视觉建模", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "We propose a simple pairwise sigmoid loss for imagetext pre-training. Unlike standard contrastive learning with softmax normalization, the sigmoid loss operates solely on image-text pairs and does not require a global view of the pairwise similarities for normalization. The sigmoid loss simultaneously allows further scaling up the batch size, while also performing better at smaller batch sizes. With only four TPUv4 chips, we can train a Base CLIP model at 4k batch size and a Large LiT model at 20k batch size, the latter achieves 84.5% ImageNet zero-shot accuracy in two days. This disentanglement of the batch size from the loss further allows us to study the impact of examples vs pairs and negative to positive ratio. Finally, we push the batch size to the extreme, up to one million, and find that the benefits of growing batch size quickly diminish, with a more reasonable batch size of 32k being sufficient. We hope our research motivates further explorations in improving the quality and efficiency of language-image pre-training.", "motivation": "Unlike standard contrastive learning with softmax normalization, the sigmoid loss operates solely on image-text pairs and does not require a global view of the pairwise similarities for normalization.", "implementation": "We propose a simple pairwise sigmoid loss for imagetext pre-training.", "application": "With only four TPUv4 chips, we can train a Base CLIP model at 4k batch size and a Large LiT model at 20k batch size, the latter achieves 84.5% ImageNet zero-shot accuracy in two days.", "problem": "Unlike standard contrastive learning with softmax normalization, the sigmoid loss operates solely on image-text pairs and does not require a global view of the pairwise similarities for normalization.", "contribution": "We propose a simple pairwise sigmoid loss for imagetext pre-training.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "Sigmoid Loss for Language Image Pre-Training 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Vision-Language / Foundation Models 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 3,004；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何降低标注和算力成本，同时保持可控性、可解释性和领域泛化。", "directionNote": "该方向的高引用工作集中体现了 Vision-Language / Foundation Models 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "ICCV", "Vision-Language / Foundation Models" ], "code": { "available": false, "url": "" }, "url": "https://www.semanticscholar.org/paper/35aba190f28b5c39df333c06ca21f46bd4845eba", "updated": "2026-05-20", "topCitedRank": 95, "citationCount": 3004, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/35aba190f28b5c39df333c06ca21f46bd4845eba", "doi": "10.1109/ICCV51070.2023.01100", "arxiv": "2303.15343" }, { "title": "FaceForensics++: Learning to Detect Manipulated Facial Images", "authors": "Andreas Rössler, D. Cozzolino, L. Verdoliva et al.", "venue": "ICCV", "year": 2019, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "The rapid progress in synthetic image generation and manipulation has now come to a point where it raises significant concerns for the implications towards society. At best, this leads to a loss of trust in digital content, but could potentially cause further harm by spreading false information or fake news. This paper examines the realism of state-of-the-art image manipulations, and how difficult it is to detect them, either automatically or by humans. To standardize the evaluation of detection methods, we propose an automated benchmark for facial manipulation detection. In particular, the benchmark is based on Deep-Fakes, Face2Face, FaceSwap and NeuralTextures as prominent representatives for facial manipulations at random compression level and size. The benchmark is publicly available and contains a hidden test set as well as a database of over 1.8 million manipulated images. This dataset is over an order of magnitude larger than comparable, publicly available, forgery datasets. Based on this data, we performed a thorough analysis of data-driven forgery detectors. We show that the use of additional domain-specific knowledge improves forgery detection to unprecedented accuracy, even in the presence of strong compression, and clearly outperforms human observers.", "motivation": "This paper examines the realism of state-of-the-art image manipulations, and how difficult it is to detect them, either automatically or by humans.", "implementation": "At best, this leads to a loss of trust in digital content, but could potentially cause further harm by spreading false information or fake news.", "application": "The rapid progress in synthetic image generation and manipulation has now come to a point where it raises significant concerns for the implications towards society.", "problem": "This paper examines the realism of state-of-the-art image manipulations, and how difficult it is to detect them, either automatically or by humans.", "contribution": "At best, this leads to a loss of trust in digital content, but could potentially cause further harm by spreading false information or fake news.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "FaceForensics++: Learning to Detect Manipulated Facial Images 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 3,002；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "ICCV", "Datasets / Benchmarks" ], "code": { "available": false, "url": "" }, "url": "https://arxiv.org/pdf/1901.08971", "updated": "2026-05-20", "topCitedRank": 96, "citationCount": 3002, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/b4f8c1353aa2d88cacfaef1b3afba74dbf427d89", "doi": "10.1109/ICCV.2019.00009", "arxiv": "1901.08971" }, { "title": "CCNet: Criss-Cross Attention for Semantic Segmentation", "authors": "Zilong Huang, Xinggang Wang, Lichao Huang et al.", "venue": "ICCV", "year": 2018, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "Full-image dependencies provide useful contextual information to benefit visual understanding problems. In this work, we propose a Criss-Cross Network (CCNet) for obtaining such contextual information in a more effective and efficient way. Concretely, for each pixel, a novel criss-cross attention module in CCNet harvests the contextual information of all the pixels on its criss-cross path. By taking a further recurrent operation, each pixel can finally capture the full-image dependencies from all pixels. Overall, CCNet is with the following merits: 1) GPU memory friendly. Compared with the non-local block, the proposed recurrent criss-cross attention module requires 11x less GPU memory usage. 2) High computational efficiency. The recurrent criss-cross attention significantly reduces FLOPs by about 85% of the non-local block in computing full-image dependencies. 3) The state-of-the-art performance. We conduct extensive experiments on popular semantic segmentation benchmarks including Cityscapes, ADE20K, and instance segmentation benchmark COCO. In particular, our CCNet achieves the mIoU score of 81.4 and 45.22 on Cityscapes test set and ADE20K validation set, respectively, which are the new state-of-the-art results. The source code is available at https://github.com/speedinghzl/CCNet.", "motivation": "Compared with the non-local block, the proposed recurrent criss-cross attention module requires 11x less GPU memory usage.", "implementation": "In this work, we propose a Criss-Cross Network (CCNet) for obtaining such contextual information in a more effective and efficient way.", "application": "We conduct extensive experiments on popular semantic segmentation benchmarks including Cityscapes, ADE20K, and instance segmentation benchmark COCO.", "problem": "Compared with the non-local block, the proposed recurrent criss-cross attention module requires 11x less GPU memory usage.", "contribution": "In this work, we propose a Criss-Cross Network (CCNet) for obtaining such contextual information in a more effective and efficient way.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "CCNet: Criss-Cross Attention for Semantic Segmentation 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 2,982；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "ICCV", "Datasets / Benchmarks" ], "code": { "available": true, "url": "https://github.com/speedinghzl/CCNet" }, "url": "https://arxiv.org/pdf/1811.11721", "updated": "2026-05-20", "topCitedRank": 97, "citationCount": 2982, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/5132500b23d2da47129b3f4f68dd30947a29e502", "doi": "10.1109/ICCV.2019.00069", "arxiv": "1811.11721" }, { "title": "ViViT: A Video Vision Transformer", "authors": "Anurag Arnab, Mostafa Dehghani, G. Heigold et al.", "venue": "ICCV", "year": 2021, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "We present pure-transformer based models for video classification, drawing upon the recent success of such models in image classification. Our model extracts spatiotemporal tokens from the input video, which are then encoded by a series of transformer layers. In order to handle the long sequences of tokens encountered in video, we propose several, efficient variants of our model which factorise the spatial- and temporal-dimensions of the input. Although transformer-based models are known to only be effective when large training datasets are available, we show how we can effectively regularise the model during training and leverage pretrained image models to be able to train on comparatively small datasets. We conduct thorough ablation studies, and achieve state-of-the-art results on multiple video classification benchmarks including Kinetics 400 and 600, Epic Kitchens, Something-Something v2 and Moments in Time, outperforming prior methods based on deep 3D convolutional networks.", "motivation": "Although transformer-based models are known to only be effective when large training datasets are available, we show how we can effectively regularise the model during training and leverage pretrained image models to be able to train on comparatively small datasets.", "implementation": "Our model extracts spatiotemporal tokens from the input video, which are then encoded by a series of transformer layers.", "application": "We present pure-transformer based models for video classification, drawing upon the recent success of such models in image classification.", "problem": "Although transformer-based models are known to only be effective when large training datasets are available, we show how we can effectively regularise the model during training and leverage pretrained image models to be able to train on comparatively small datasets.", "contribution": "Our model extracts spatiotemporal tokens from the input video, which are then encoded by a series of transformer layers.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "ViViT: A Video Vision Transformer 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 2,966；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "ICCV", "Datasets / Benchmarks" ], "code": { "available": false, "url": "" }, "url": "https://arxiv.org/pdf/2103.15691", "updated": "2026-05-20", "topCitedRank": 98, "citationCount": 2966, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/b6382a7351c0c595f91472ac71d3b2d87b3c4844", "doi": "10.1109/ICCV48922.2021.00676", "arxiv": "2103.15691" }, { "title": "InstructPix2Pix: Learning to Follow Image Editing Instructions", "authors": "Tim Brooks, Aleksander Holynski, Alexei A. Efros", "venue": "CVPR", "year": 2022, "topic": "Vision-Language / Foundation Models", "modality": "Image-text / foundation model", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "视觉-语言理解、开放词表识别或基础模型适配", "method": "大规模预训练、多模态对齐或提示式视觉建模", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "We propose a method for editing images from human instructions: given an input image and a written instruction that tells the model what to do, our model follows these instructions to edit the image. To obtain training data for this problem, we combine the knowledge of two large pretrained models—a language model (GPT-3) and a text-to-image model (Stable Diffusion)—to generate a large dataset of image editing examples. Our conditional diffusion model, InstructPix2Pix, is trained on our generated data, and generalizes to real images and user-written instructions at inference time. Since it performs edits in the forward pass and does not require per-example fine-tuning or inversion, our model edits images quickly, in a matter of seconds. We show compelling editing results for a diverse collection of input images and written instructions.", "motivation": "To obtain training data for this problem, we combine the knowledge of two large pretrained models—a language model (GPT-3) and a text-to-image model (Stable Diffusion)—to generate a large dataset of image editing examples.", "implementation": "We propose a method for editing images from human instructions: given an input image and a written instruction that tells the model what to do, our model follows these instructions to edit the image.", "application": "We show compelling editing results for a diverse collection of input images and written instructions.", "problem": "To obtain training data for this problem, we combine the knowledge of two large pretrained models—a language model (GPT-3) and a text-to-image model (Stable Diffusion)—to generate a large dataset of image editing examples.", "contribution": "We propose a method for editing images from human instructions: given an input image and a written instruction that tells the model what to do, our model follows these instructions to edit the image.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "InstructPix2Pix: Learning to Follow Image Editing Instructions 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Vision-Language / Foundation Models 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 2,940；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何降低标注和算力成本，同时保持可控性、可解释性和领域泛化。", "directionNote": "该方向的高引用工作集中体现了 Vision-Language / Foundation Models 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Vision-Language / Foundation Models" ], "code": { "available": false, "url": "" }, "url": "https://arxiv.org/pdf/2211.09800", "updated": "2026-05-20", "topCitedRank": 99, "citationCount": 2940, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/a2d2bbe4c542173662a444b33b76c66992697830", "doi": "10.1109/CVPR52729.2023.01764", "arxiv": "2211.09800" }, { "title": "NTU RGB+D: A Large Scale Dataset for 3D Human Activity Analysis", "authors": "Amir Shahroudy, Jun Liu, T. Ng et al.", "venue": "CVPR", "year": 2016, "topic": "Datasets / Benchmarks", "modality": "Benchmark dataset", "status": "CV 顶会高引 Top 100", "priority": "Medium", "task": "构建大规模数据集或评测基准", "method": "数据采集、标注协议和标准化评测任务", "hardware": "常规视觉数据与公开基准", "dataset": "见论文实验设置", "abstract": "Recent approaches in depth-based human activity analysis achieved outstanding performance and proved the effectiveness of 3D representation for classification of action classes. Currently available depth-based and RGB+Dbased action recognition benchmarks have a number of limitations, including the lack of training samples, distinct class labels, camera views and variety of subjects. In this paper we introduce a large-scale dataset for RGB+D human action recognition with more than 56 thousand video samples and 4 million frames, collected from 40 distinct subjects. Our dataset contains 60 different action classes including daily, mutual, and health-related actions. In addition, we propose a new recurrent neural network structure to model the long-term temporal correlation of the features for each body part, and utilize them for better action classification. Experimental results show the advantages of applying deep learning methods over state-of-the-art handcrafted features on the suggested cross-subject and cross-view evaluation criteria for our dataset. The introduction of this large scale dataset will enable the community to apply, develop and adapt various data-hungry learning techniques for the task of depth-based and RGB+D-based human activity analysis.", "motivation": "Currently available depth-based and RGB+Dbased action recognition benchmarks have a number of limitations, including the lack of training samples, distinct class labels, camera views and variety of subjects.", "implementation": "In this paper we introduce a large-scale dataset for RGB+D human action recognition with more than 56 thousand video samples and 4 million frames, collected from 40 distinct subjects.", "application": "Recent approaches in depth-based human activity analysis achieved outstanding performance and proved the effectiveness of 3D representation for classification of action classes.", "problem": "Currently available depth-based and RGB+Dbased action recognition benchmarks have a number of limitations, including the lack of training samples, distinct class labels, camera views and variety of subjects.", "contribution": "In this paper we introduce a large-scale dataset for RGB+D human action recognition with more than 56 thousand video samples and 4 million frames, collected from 40 distinct subjects.", "limitation": "高引用代表历史影响力，不等于当前最优；使用时需要结合后续工作、任务设定和数据偏差判断边界。", "insight": "NTU RGB+D: A Large Scale Dataset for 3D Human Activity Analysis 是近十年 CVPR/ECCV/ICCV 高引用代表作，适合作为 Datasets / Benchmarks 方向的入口论文。", "whyFollow": "Semantic Scholar 抓取时引用数约 2,930；适合用于梳理近十年计算机视觉主线方法和基准演化。", "openQuestion": "如何避免数据偏差，并让基准持续反映真实应用需求。", "directionNote": "该方向的高引用工作集中体现了 Datasets / Benchmarks 从方法提出到标准基线的扩散过程。", "tags": [ "CVPR/ECCV/ICCV", "top-cited", "CVPR", "Datasets / Benchmarks" ], "code": { "available": false, "url": "" }, "url": "http://arxiv.org/pdf/1604.02808", "updated": "2026-05-20", "topCitedRank": 100, "citationCount": 2930, "citationSource": "Semantic Scholar", "citationRetrieved": "2026-05-20", "semanticScholarUrl": "https://www.semanticscholar.org/paper/091e4d3c85dc0a8212afea875cd3b162d273d46b", "doi": "10.1109/CVPR.2016.115", "arxiv": "1604.02808" }, { "title": "VGGT: Visual Geometry Grounded Transformer", "authors": "Jianyuan Wang, Minghao Chen, Nikita Karaev, Andrea Vedaldi, Christian Rupprecht, David Novotny", "venue": "CVPR", "year": 2025, "topic": "3D / Geometry / Reconstruction", "modality": "3D / Geometry / Reconstruction", "status": "CV 顶会 Best Paper 合集", "priority": "High", "task": "3D / Geometry / Reconstruction", "method": "VGGT: Visual Geometry Grounded Transformer introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "VGGT: Visual Geometry Grounded Transformer is an award-winning computer vision paper.", "motivation": "VGGT: Visual Geometry Grounded Transformer targets a central problem recognized by the conference awards committee.", "implementation": "VGGT: Visual Geometry Grounded Transformer introduces a method or system selected as award-winning work.", "application": "VGGT: Visual Geometry Grounded Transformer is relevant to its conference track and downstream computer vision applications.", "problem": "VGGT: Visual Geometry Grounded Transformer targets a central problem recognized by the conference awards committee.", "contribution": "VGGT: Visual Geometry Grounded Transformer introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "CVPR 2025 Best Paper: VGGT: Visual Geometry Grounded Transformer is an award-winning computer vision paper.", "whyFollow": "CVPR 2025 Best Paper，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "VGGT: Visual Geometry Grounded Transformer is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "CVPR", "Best Paper", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "CVPR", "year": 2025, "award": "Best Paper", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "Neural Inverse Rendering from Propagating Light", "authors": "Anagh Malik, Benjamin Attal, Andrew Xie, Matthew O'Toole, David B. Lindell", "venue": "CVPR", "year": 2025, "topic": "Computational Imaging", "modality": "Computational Imaging", "status": "CV 顶会 Best Paper 合集", "priority": "High", "task": "Computational Imaging", "method": "Neural Inverse Rendering from Propagating Light introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "Neural Inverse Rendering from Propagating Light is an award-winning computer vision paper.", "motivation": "Neural Inverse Rendering from Propagating Light targets a central problem recognized by the conference awards committee.", "implementation": "Neural Inverse Rendering from Propagating Light introduces a method or system selected as award-winning work.", "application": "Neural Inverse Rendering from Propagating Light is relevant to its conference track and downstream computer vision applications.", "problem": "Neural Inverse Rendering from Propagating Light targets a central problem recognized by the conference awards committee.", "contribution": "Neural Inverse Rendering from Propagating Light introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "CVPR 2025 Best Student Paper: Neural Inverse Rendering from Propagating Light is an award-winning computer vision paper.", "whyFollow": "CVPR 2025 Best Student Paper，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "Neural Inverse Rendering from Propagating Light is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "CVPR", "Best Student", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "CVPR", "year": 2025, "award": "Best Student Paper", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "MegaSaM: Accurate, Fast and Robust Structure and Motion from Casual Dynamic Videos", "authors": "Zhengqi Li, Richard Tucker, Forrester Cole, Qianqian Wang, Linyi Jin, Vickie Ye, Angjoo Kanazawa, Aleksander Holynski, Noah Snavely", "venue": "CVPR", "year": 2025, "topic": "3D / Geometry / Reconstruction", "modality": "3D / Geometry / Reconstruction", "status": "CV 顶会 Best Paper 合集", "priority": "Medium", "task": "3D / Geometry / Reconstruction", "method": "MegaSaM: Accurate, Fast and Robust Structure and Motion from Casual Dynamic Videos introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "MegaSaM: Accurate, Fast and Robust Structure and Motion from Casual Dynamic Videos is an award-winning computer vision paper.", "motivation": "MegaSaM: Accurate, Fast and Robust Structure and Motion from Casual Dynamic Videos targets a central problem recognized by the conference awards committee.", "implementation": "MegaSaM: Accurate, Fast and Robust Structure and Motion from Casual Dynamic Videos introduces a method or system selected as award-winning work.", "application": "MegaSaM: Accurate, Fast and Robust Structure and Motion from Casual Dynamic Videos is relevant to its conference track and downstream computer vision applications.", "problem": "MegaSaM: Accurate, Fast and Robust Structure and Motion from Casual Dynamic Videos targets a central problem recognized by the conference awards committee.", "contribution": "MegaSaM: Accurate, Fast and Robust Structure and Motion from Casual Dynamic Videos introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "CVPR 2025 Honorable Mention: MegaSaM: Accurate, Fast and Robust Structure and Motion from Casual Dynamic Videos is an award-winning computer vision paper.", "whyFollow": "CVPR 2025 Honorable Mention，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "MegaSaM: Accurate, Fast and Robust Structure and Motion from Casual Dynamic Videos is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "CVPR", "Honorable Mention", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "CVPR", "year": 2025, "award": "Honorable Mention", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "Navigation World Models", "authors": "Amir Bar, Gaoyue Zhou, Danny Tran, Trevor Darrell, Yann LeCun", "venue": "CVPR", "year": 2025, "topic": "Video / Embodied Perception", "modality": "Video / Embodied Perception", "status": "CV 顶会 Best Paper 合集", "priority": "Medium", "task": "Video / Embodied Perception", "method": "Navigation World Models introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "Navigation World Models is an award-winning computer vision paper.", "motivation": "Navigation World Models targets a central problem recognized by the conference awards committee.", "implementation": "Navigation World Models introduces a method or system selected as award-winning work.", "application": "Navigation World Models is relevant to its conference track and downstream computer vision applications.", "problem": "Navigation World Models targets a central problem recognized by the conference awards committee.", "contribution": "Navigation World Models introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "CVPR 2025 Honorable Mention: Navigation World Models is an award-winning computer vision paper.", "whyFollow": "CVPR 2025 Honorable Mention，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "Navigation World Models is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "CVPR", "Honorable Mention", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "CVPR", "year": 2025, "award": "Honorable Mention", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "Molmo and PixMo: Open Weights and Open Data for State-of-the-Art Vision-Language Models", "authors": "Matt Deitke et al.", "venue": "CVPR", "year": 2025, "topic": "Vision-Language / Foundation Models", "modality": "Vision-Language / Foundation Models", "status": "CV 顶会 Best Paper 合集", "priority": "Medium", "task": "Vision-Language / Foundation Models", "method": "Molmo and PixMo: Open Weights and Open Data for State-of-the-Art Vision-Language Models introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "Molmo and PixMo: Open Weights and Open Data for State-of-the-Art Vision-Language Models is an award-winning computer vision paper.", "motivation": "Molmo and PixMo: Open Weights and Open Data for State-of-the-Art Vision-Language Models targets a central problem recognized by the conference awards committee.", "implementation": "Molmo and PixMo: Open Weights and Open Data for State-of-the-Art Vision-Language Models introduces a method or system selected as award-winning work.", "application": "Molmo and PixMo: Open Weights and Open Data for State-of-the-Art Vision-Language Models is relevant to its conference track and downstream computer vision applications.", "problem": "Molmo and PixMo: Open Weights and Open Data for State-of-the-Art Vision-Language Models targets a central problem recognized by the conference awards committee.", "contribution": "Molmo and PixMo: Open Weights and Open Data for State-of-the-Art Vision-Language Models introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "CVPR 2025 Honorable Mention: Molmo and PixMo: Open Weights and Open Data for State-of-the-Art Vision-Language Models is an award-winning computer vision paper.", "whyFollow": "CVPR 2025 Honorable Mention，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "Molmo and PixMo: Open Weights and Open Data for State-of-the-Art Vision-Language Models is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "CVPR", "Honorable Mention", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "CVPR", "year": 2025, "award": "Honorable Mention", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "3D Student Splatting and Scooping", "authors": "Jialin Zhu, Jiangbei Yue, Feixiang He, He Wang", "venue": "CVPR", "year": 2025, "topic": "3D / Geometry / Reconstruction", "modality": "3D / Geometry / Reconstruction", "status": "CV 顶会 Best Paper 合集", "priority": "Medium", "task": "3D / Geometry / Reconstruction", "method": "3D Student Splatting and Scooping introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "3D Student Splatting and Scooping is an award-winning computer vision paper.", "motivation": "3D Student Splatting and Scooping targets a central problem recognized by the conference awards committee.", "implementation": "3D Student Splatting and Scooping introduces a method or system selected as award-winning work.", "application": "3D Student Splatting and Scooping is relevant to its conference track and downstream computer vision applications.", "problem": "3D Student Splatting and Scooping targets a central problem recognized by the conference awards committee.", "contribution": "3D Student Splatting and Scooping introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "CVPR 2025 Honorable Mention: 3D Student Splatting and Scooping is an award-winning computer vision paper.", "whyFollow": "CVPR 2025 Honorable Mention，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "3D Student Splatting and Scooping is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "CVPR", "Honorable Mention", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "CVPR", "year": 2025, "award": "Honorable Mention", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "Generative Multimodal Pretraining with Discrete Diffusion Timestep Tokens", "authors": "Kaihang Pan, Wang Lin, Zhongqi Yue, Tenglong Ao, Liyu Jia, Wei Zhao, Juncheng Li, Siliang Tang, Hanwang Zhang", "venue": "CVPR", "year": 2025, "topic": "Generative Vision", "modality": "Generative Vision", "status": "CV 顶会 Best Paper 合集", "priority": "Medium", "task": "Generative Vision", "method": "Generative Multimodal Pretraining with Discrete Diffusion Timestep Tokens introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "Generative Multimodal Pretraining with Discrete Diffusion Timestep Tokens is an award-winning computer vision paper.", "motivation": "Generative Multimodal Pretraining with Discrete Diffusion Timestep Tokens targets a central problem recognized by the conference awards committee.", "implementation": "Generative Multimodal Pretraining with Discrete Diffusion Timestep Tokens introduces a method or system selected as award-winning work.", "application": "Generative Multimodal Pretraining with Discrete Diffusion Timestep Tokens is relevant to its conference track and downstream computer vision applications.", "problem": "Generative Multimodal Pretraining with Discrete Diffusion Timestep Tokens targets a central problem recognized by the conference awards committee.", "contribution": "Generative Multimodal Pretraining with Discrete Diffusion Timestep Tokens introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "CVPR 2025 Honorable Mention: Generative Multimodal Pretraining with Discrete Diffusion Timestep Tokens is an award-winning computer vision paper.", "whyFollow": "CVPR 2025 Honorable Mention，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "Generative Multimodal Pretraining with Discrete Diffusion Timestep Tokens is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "CVPR", "Honorable Mention", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "CVPR", "year": 2025, "award": "Honorable Mention", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "Rich Human Feedback for Text-to-Image Generation", "authors": "Youwei Liang, Junfeng He, Gang Li, Peizhao Li, Arseniy Klimovskiy, Nicholas Carolan, Jiao Sun, Jordi Pont-Tuset, Sarah Young, Feng Yang, Junjie Ke, Krishnamurthy Dj Dvijotham, Katherine M. Collins, Yiwen Luo, Yang Li, Kai J. Kohlhoff, Deepak Ramachandran, Vidhya Navalpakkam", "venue": "CVPR", "year": 2024, "topic": "Generative Vision", "modality": "Generative Vision", "status": "CV 顶会 Best Paper 合集", "priority": "High", "task": "Generative Vision", "method": "Rich Human Feedback for Text-to-Image Generation introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "Rich Human Feedback for Text-to-Image Generation is an award-winning computer vision paper.", "motivation": "Rich Human Feedback for Text-to-Image Generation targets a central problem recognized by the conference awards committee.", "implementation": "Rich Human Feedback for Text-to-Image Generation introduces a method or system selected as award-winning work.", "application": "Rich Human Feedback for Text-to-Image Generation is relevant to its conference track and downstream computer vision applications.", "problem": "Rich Human Feedback for Text-to-Image Generation targets a central problem recognized by the conference awards committee.", "contribution": "Rich Human Feedback for Text-to-Image Generation introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "CVPR 2024 Best Paper: Rich Human Feedback for Text-to-Image Generation is an award-winning computer vision paper.", "whyFollow": "CVPR 2024 Best Paper，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "Rich Human Feedback for Text-to-Image Generation is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "CVPR", "Best Paper", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "CVPR", "year": 2024, "award": "Best Paper", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "Generative Image Dynamics", "authors": "Zhengqi Li, Richard Tucker, Noah Snavely, Aleksander Holynski", "venue": "CVPR", "year": 2024, "topic": "Generative Vision", "modality": "Generative Vision", "status": "CV 顶会 Best Paper 合集", "priority": "High", "task": "Generative Vision", "method": "Generative Image Dynamics introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "Generative Image Dynamics is an award-winning computer vision paper.", "motivation": "Generative Image Dynamics targets a central problem recognized by the conference awards committee.", "implementation": "Generative Image Dynamics introduces a method or system selected as award-winning work.", "application": "Generative Image Dynamics is relevant to its conference track and downstream computer vision applications.", "problem": "Generative Image Dynamics targets a central problem recognized by the conference awards committee.", "contribution": "Generative Image Dynamics introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "CVPR 2024 Best Paper: Generative Image Dynamics is an award-winning computer vision paper.", "whyFollow": "CVPR 2024 Best Paper，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "Generative Image Dynamics is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "CVPR", "Best Paper", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "CVPR", "year": 2024, "award": "Best Paper", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "BioCLIP: A Vision Foundation Model for the Tree of Life", "authors": "Samuel Stevens, Jiaman Wu, Matthew J. Thompson, Elizabeth G. Campolongo, Connor H. Song, David E. Carlyn, Li Dong, Wasila M. Dahdul, Charles Stewart, Tanya Berger-Wolf, Wei-Lun Chao, Yu Su", "venue": "CVPR", "year": 2024, "topic": "Vision-Language / Foundation Models", "modality": "Vision-Language / Foundation Models", "status": "CV 顶会 Best Paper 合集", "priority": "High", "task": "Vision-Language / Foundation Models", "method": "BioCLIP: A Vision Foundation Model for the Tree of Life introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "BioCLIP: A Vision Foundation Model for the Tree of Life is an award-winning computer vision paper.", "motivation": "BioCLIP: A Vision Foundation Model for the Tree of Life targets a central problem recognized by the conference awards committee.", "implementation": "BioCLIP: A Vision Foundation Model for the Tree of Life introduces a method or system selected as award-winning work.", "application": "BioCLIP: A Vision Foundation Model for the Tree of Life is relevant to its conference track and downstream computer vision applications.", "problem": "BioCLIP: A Vision Foundation Model for the Tree of Life targets a central problem recognized by the conference awards committee.", "contribution": "BioCLIP: A Vision Foundation Model for the Tree of Life introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "CVPR 2024 Best Student Paper: BioCLIP: A Vision Foundation Model for the Tree of Life is an award-winning computer vision paper.", "whyFollow": "CVPR 2024 Best Student Paper，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "BioCLIP: A Vision Foundation Model for the Tree of Life is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "CVPR", "Best Student", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "CVPR", "year": 2024, "award": "Best Student Paper", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "Mip-Splatting: Alias-free 3D Gaussian Splatting", "authors": "Zehao Yu, Anpei Chen, Binbin Huang, Torsten Sattler, Andreas Geiger", "venue": "CVPR", "year": 2024, "topic": "3D / Geometry / Reconstruction", "modality": "3D / Geometry / Reconstruction", "status": "CV 顶会 Best Paper 合集", "priority": "High", "task": "3D / Geometry / Reconstruction", "method": "Mip-Splatting: Alias-free 3D Gaussian Splatting introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "Mip-Splatting: Alias-free 3D Gaussian Splatting is an award-winning computer vision paper.", "motivation": "Mip-Splatting: Alias-free 3D Gaussian Splatting targets a central problem recognized by the conference awards committee.", "implementation": "Mip-Splatting: Alias-free 3D Gaussian Splatting introduces a method or system selected as award-winning work.", "application": "Mip-Splatting: Alias-free 3D Gaussian Splatting is relevant to its conference track and downstream computer vision applications.", "problem": "Mip-Splatting: Alias-free 3D Gaussian Splatting targets a central problem recognized by the conference awards committee.", "contribution": "Mip-Splatting: Alias-free 3D Gaussian Splatting introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "CVPR 2024 Best Student Paper: Mip-Splatting: Alias-free 3D Gaussian Splatting is an award-winning computer vision paper.", "whyFollow": "CVPR 2024 Best Student Paper，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "Mip-Splatting: Alias-free 3D Gaussian Splatting is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "CVPR", "Best Student", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "CVPR", "year": 2024, "award": "Best Student Paper", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "pixelSplat: 3D Gaussian Splats from Image Pairs for Scalable Generalizable 3D Reconstruction", "authors": "David Charatan, Sizhe Li, Andrea Tagliasacchi, Vincent Sitzmann", "venue": "CVPR", "year": 2024, "topic": "3D / Geometry / Reconstruction", "modality": "3D / Geometry / Reconstruction", "status": "CV 顶会 Best Paper 合集", "priority": "Medium", "task": "3D / Geometry / Reconstruction", "method": "pixelSplat: 3D Gaussian Splats from Image Pairs for Scalable Generalizable 3D Reconstruction introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "pixelSplat: 3D Gaussian Splats from Image Pairs for Scalable Generalizable 3D Reconstruction is an award-winning computer vision paper.", "motivation": "pixelSplat: 3D Gaussian Splats from Image Pairs for Scalable Generalizable 3D Reconstruction targets a central problem recognized by the conference awards committee.", "implementation": "pixelSplat: 3D Gaussian Splats from Image Pairs for Scalable Generalizable 3D Reconstruction introduces a method or system selected as award-winning work.", "application": "pixelSplat: 3D Gaussian Splats from Image Pairs for Scalable Generalizable 3D Reconstruction is relevant to its conference track and downstream computer vision applications.", "problem": "pixelSplat: 3D Gaussian Splats from Image Pairs for Scalable Generalizable 3D Reconstruction targets a central problem recognized by the conference awards committee.", "contribution": "pixelSplat: 3D Gaussian Splats from Image Pairs for Scalable Generalizable 3D Reconstruction introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "CVPR 2024 Honorable Mention: pixelSplat: 3D Gaussian Splats from Image Pairs for Scalable Generalizable 3D Reconstruction is an award-winning computer vision paper.", "whyFollow": "CVPR 2024 Honorable Mention，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "pixelSplat: 3D Gaussian Splats from Image Pairs for Scalable Generalizable 3D Reconstruction is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "CVPR", "Honorable Mention", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "CVPR", "year": 2024, "award": "Honorable Mention", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "EventPS: Real-Time Photometric Stereo Using an Event Camera", "authors": "Boxin Yu, Jie Ren, Jiaying Han, Feng Wang, Boxin Shi", "venue": "CVPR", "year": 2024, "topic": "Computational Imaging", "modality": "Computational Imaging", "status": "CV 顶会 Best Paper 合集", "priority": "Medium", "task": "Computational Imaging", "method": "EventPS: Real-Time Photometric Stereo Using an Event Camera introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "EventPS: Real-Time Photometric Stereo Using an Event Camera is an award-winning computer vision paper.", "motivation": "EventPS: Real-Time Photometric Stereo Using an Event Camera targets a central problem recognized by the conference awards committee.", "implementation": "EventPS: Real-Time Photometric Stereo Using an Event Camera introduces a method or system selected as award-winning work.", "application": "EventPS: Real-Time Photometric Stereo Using an Event Camera is relevant to its conference track and downstream computer vision applications.", "problem": "EventPS: Real-Time Photometric Stereo Using an Event Camera targets a central problem recognized by the conference awards committee.", "contribution": "EventPS: Real-Time Photometric Stereo Using an Event Camera introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "CVPR 2024 Honorable Mention: EventPS: Real-Time Photometric Stereo Using an Event Camera is an award-winning computer vision paper.", "whyFollow": "CVPR 2024 Honorable Mention，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "EventPS: Real-Time Photometric Stereo Using an Event Camera is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "CVPR", "Honorable Mention", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "CVPR", "year": 2024, "award": "Honorable Mention", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "Comparing the Decision-Making Mechanisms by Transformers and CNNs via Explanation Methods", "authors": "Mingyu Jiang, Sara Khorram, Fuxin Li", "venue": "CVPR", "year": 2024, "topic": "Vision Backbones", "modality": "Vision Backbones", "status": "CV 顶会 Best Paper 合集", "priority": "Medium", "task": "Vision Backbones", "method": "Comparing the Decision-Making Mechanisms by Transformers and CNNs via Explanation Methods introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "Comparing the Decision-Making Mechanisms by Transformers and CNNs via Explanation Methods is an award-winning computer vision paper.", "motivation": "Comparing the Decision-Making Mechanisms by Transformers and CNNs via Explanation Methods targets a central problem recognized by the conference awards committee.", "implementation": "Comparing the Decision-Making Mechanisms by Transformers and CNNs via Explanation Methods introduces a method or system selected as award-winning work.", "application": "Comparing the Decision-Making Mechanisms by Transformers and CNNs via Explanation Methods is relevant to its conference track and downstream computer vision applications.", "problem": "Comparing the Decision-Making Mechanisms by Transformers and CNNs via Explanation Methods targets a central problem recognized by the conference awards committee.", "contribution": "Comparing the Decision-Making Mechanisms by Transformers and CNNs via Explanation Methods introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "CVPR 2024 Honorable Mention: Comparing the Decision-Making Mechanisms by Transformers and CNNs via Explanation Methods is an award-winning computer vision paper.", "whyFollow": "CVPR 2024 Honorable Mention，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "Comparing the Decision-Making Mechanisms by Transformers and CNNs via Explanation Methods is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "CVPR", "Honorable Mention", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "CVPR", "year": 2024, "award": "Honorable Mention", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "Objects as Volumes: A Stochastic Geometry View of Opaque Solids", "authors": "Bailey Miller, Hsueh-Ti Derek Liu Chen, I-Chao Shen, Ioannis Gkioulekas", "venue": "CVPR", "year": 2024, "topic": "3D / Geometry / Reconstruction", "modality": "3D / Geometry / Reconstruction", "status": "CV 顶会 Best Paper 合集", "priority": "Medium", "task": "3D / Geometry / Reconstruction", "method": "Objects as Volumes: A Stochastic Geometry View of Opaque Solids introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "Objects as Volumes: A Stochastic Geometry View of Opaque Solids is an award-winning computer vision paper.", "motivation": "Objects as Volumes: A Stochastic Geometry View of Opaque Solids targets a central problem recognized by the conference awards committee.", "implementation": "Objects as Volumes: A Stochastic Geometry View of Opaque Solids introduces a method or system selected as award-winning work.", "application": "Objects as Volumes: A Stochastic Geometry View of Opaque Solids is relevant to its conference track and downstream computer vision applications.", "problem": "Objects as Volumes: A Stochastic Geometry View of Opaque Solids targets a central problem recognized by the conference awards committee.", "contribution": "Objects as Volumes: A Stochastic Geometry View of Opaque Solids introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "CVPR 2024 Honorable Mention: Objects as Volumes: A Stochastic Geometry View of Opaque Solids is an award-winning computer vision paper.", "whyFollow": "CVPR 2024 Honorable Mention，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "Objects as Volumes: A Stochastic Geometry View of Opaque Solids is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "CVPR", "Honorable Mention", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "CVPR", "year": 2024, "award": "Honorable Mention", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "Image Processing GNN: Breaking Rigidity in Super-Resolution", "authors": "Yuechen Tian, Hanting Chen, Chao Xu, Yunhe Wang", "venue": "CVPR", "year": 2024, "topic": "Award-Winning Computer Vision", "modality": "Award-Winning Computer Vision", "status": "CV 顶会 Best Paper 合集", "priority": "Medium", "task": "Award-Winning Computer Vision", "method": "Image Processing GNN: Breaking Rigidity in Super-Resolution introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "Image Processing GNN: Breaking Rigidity in Super-Resolution is an award-winning computer vision paper.", "motivation": "Image Processing GNN: Breaking Rigidity in Super-Resolution targets a central problem recognized by the conference awards committee.", "implementation": "Image Processing GNN: Breaking Rigidity in Super-Resolution introduces a method or system selected as award-winning work.", "application": "Image Processing GNN: Breaking Rigidity in Super-Resolution is relevant to its conference track and downstream computer vision applications.", "problem": "Image Processing GNN: Breaking Rigidity in Super-Resolution targets a central problem recognized by the conference awards committee.", "contribution": "Image Processing GNN: Breaking Rigidity in Super-Resolution introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "CVPR 2024 Honorable Mention: Image Processing GNN: Breaking Rigidity in Super-Resolution is an award-winning computer vision paper.", "whyFollow": "CVPR 2024 Honorable Mention，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "Image Processing GNN: Breaking Rigidity in Super-Resolution is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "CVPR", "Honorable Mention", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "CVPR", "year": 2024, "award": "Honorable Mention", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "SpiderMatch: 3D Shape Matching with Global Optimality and Geometric Consistency", "authors": "Paul Roetzer, Florian Bernard", "venue": "CVPR", "year": 2024, "topic": "3D / Geometry / Reconstruction", "modality": "3D / Geometry / Reconstruction", "status": "CV 顶会 Best Paper 合集", "priority": "Medium", "task": "3D / Geometry / Reconstruction", "method": "SpiderMatch: 3D Shape Matching with Global Optimality and Geometric Consistency introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "SpiderMatch: 3D Shape Matching with Global Optimality and Geometric Consistency is an award-winning computer vision paper.", "motivation": "SpiderMatch: 3D Shape Matching with Global Optimality and Geometric Consistency targets a central problem recognized by the conference awards committee.", "implementation": "SpiderMatch: 3D Shape Matching with Global Optimality and Geometric Consistency introduces a method or system selected as award-winning work.", "application": "SpiderMatch: 3D Shape Matching with Global Optimality and Geometric Consistency is relevant to its conference track and downstream computer vision applications.", "problem": "SpiderMatch: 3D Shape Matching with Global Optimality and Geometric Consistency targets a central problem recognized by the conference awards committee.", "contribution": "SpiderMatch: 3D Shape Matching with Global Optimality and Geometric Consistency introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "CVPR 2024 Honorable Mention: SpiderMatch: 3D Shape Matching with Global Optimality and Geometric Consistency is an award-winning computer vision paper.", "whyFollow": "CVPR 2024 Honorable Mention，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "SpiderMatch: 3D Shape Matching with Global Optimality and Geometric Consistency is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "CVPR", "Honorable Mention", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "CVPR", "year": 2024, "award": "Honorable Mention", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "Visual Programming: Compositional Visual Reasoning Without Training", "authors": "Tanmay Gupta, Aniruddha Kembhavi", "venue": "CVPR", "year": 2023, "topic": "Vision-Language / Foundation Models", "modality": "Vision-Language / Foundation Models", "status": "CV 顶会 Best Paper 合集", "priority": "High", "task": "Vision-Language / Foundation Models", "method": "Visual Programming: Compositional Visual Reasoning Without Training introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "Visual Programming: Compositional Visual Reasoning Without Training is an award-winning computer vision paper.", "motivation": "Visual Programming: Compositional Visual Reasoning Without Training targets a central problem recognized by the conference awards committee.", "implementation": "Visual Programming: Compositional Visual Reasoning Without Training introduces a method or system selected as award-winning work.", "application": "Visual Programming: Compositional Visual Reasoning Without Training is relevant to its conference track and downstream computer vision applications.", "problem": "Visual Programming: Compositional Visual Reasoning Without Training targets a central problem recognized by the conference awards committee.", "contribution": "Visual Programming: Compositional Visual Reasoning Without Training introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "CVPR 2023 Best Paper: Visual Programming: Compositional Visual Reasoning Without Training is an award-winning computer vision paper.", "whyFollow": "CVPR 2023 Best Paper，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "Visual Programming: Compositional Visual Reasoning Without Training is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "CVPR", "Best Paper", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "CVPR", "year": 2023, "award": "Best Paper", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "Planning-Oriented Autonomous Driving", "authors": "Yihan Hu, Jiazhi Yang, Li Chen, Keyu Li, Chonghao Sima, Xizhou Zhu, Siqi Chai, Senyao Du, Tianwei Lin, Wenhai Wang, Lewei Lu, Xiaosong Jia, Qiang Liu, Jifeng Dai, Yu Qiao, Hongyang Li", "venue": "CVPR", "year": 2023, "topic": "Video / Embodied Perception", "modality": "Video / Embodied Perception", "status": "CV 顶会 Best Paper 合集", "priority": "High", "task": "Video / Embodied Perception", "method": "Planning-Oriented Autonomous Driving introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "Planning-Oriented Autonomous Driving is an award-winning computer vision paper.", "motivation": "Planning-Oriented Autonomous Driving targets a central problem recognized by the conference awards committee.", "implementation": "Planning-Oriented Autonomous Driving introduces a method or system selected as award-winning work.", "application": "Planning-Oriented Autonomous Driving is relevant to its conference track and downstream computer vision applications.", "problem": "Planning-Oriented Autonomous Driving targets a central problem recognized by the conference awards committee.", "contribution": "Planning-Oriented Autonomous Driving introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "CVPR 2023 Best Paper: Planning-Oriented Autonomous Driving is an award-winning computer vision paper.", "whyFollow": "CVPR 2023 Best Paper，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "Planning-Oriented Autonomous Driving is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "CVPR", "Best Paper", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "CVPR", "year": 2023, "award": "Best Paper", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "3D Registration With Maximal Cliques", "authors": "Xiyue Zhang, Jiaqi Yang, Shikun Zhang, Yanning Zhang", "venue": "CVPR", "year": 2023, "topic": "3D / Geometry / Reconstruction", "modality": "3D / Geometry / Reconstruction", "status": "CV 顶会 Best Paper 合集", "priority": "High", "task": "3D / Geometry / Reconstruction", "method": "3D Registration With Maximal Cliques introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "3D Registration With Maximal Cliques is an award-winning computer vision paper.", "motivation": "3D Registration With Maximal Cliques targets a central problem recognized by the conference awards committee.", "implementation": "3D Registration With Maximal Cliques introduces a method or system selected as award-winning work.", "application": "3D Registration With Maximal Cliques is relevant to its conference track and downstream computer vision applications.", "problem": "3D Registration With Maximal Cliques targets a central problem recognized by the conference awards committee.", "contribution": "3D Registration With Maximal Cliques introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "CVPR 2023 Best Student Paper: 3D Registration With Maximal Cliques is an award-winning computer vision paper.", "whyFollow": "CVPR 2023 Best Student Paper，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "3D Registration With Maximal Cliques is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "CVPR", "Best Student", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "CVPR", "year": 2023, "award": "Best Student Paper", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "DynIBaR: Neural Dynamic Image-Based Rendering", "authors": "Zhengqi Li, Qianqian Wang, Forrester Cole, Richard Tucker, Noah Snavely", "venue": "CVPR", "year": 2023, "topic": "Award-Winning Computer Vision", "modality": "Award-Winning Computer Vision", "status": "CV 顶会 Best Paper 合集", "priority": "Medium", "task": "Award-Winning Computer Vision", "method": "DynIBaR: Neural Dynamic Image-Based Rendering introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "DynIBaR: Neural Dynamic Image-Based Rendering is an award-winning computer vision paper.", "motivation": "DynIBaR: Neural Dynamic Image-Based Rendering targets a central problem recognized by the conference awards committee.", "implementation": "DynIBaR: Neural Dynamic Image-Based Rendering introduces a method or system selected as award-winning work.", "application": "DynIBaR: Neural Dynamic Image-Based Rendering is relevant to its conference track and downstream computer vision applications.", "problem": "DynIBaR: Neural Dynamic Image-Based Rendering targets a central problem recognized by the conference awards committee.", "contribution": "DynIBaR: Neural Dynamic Image-Based Rendering introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "CVPR 2023 Honorable Mention: DynIBaR: Neural Dynamic Image-Based Rendering is an award-winning computer vision paper.", "whyFollow": "CVPR 2023 Honorable Mention，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "DynIBaR: Neural Dynamic Image-Based Rendering is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "CVPR", "Honorable Mention", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "CVPR", "year": 2023, "award": "Honorable Mention", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "Learning to Solve Hard Minimal Problems", "authors": "Petr Hruby, Timothy Duff, Anton Leykin, Tomas Pajdla", "venue": "CVPR", "year": 2022, "topic": "Award-Winning Computer Vision", "modality": "Award-Winning Computer Vision", "status": "CV 顶会 Best Paper 合集", "priority": "High", "task": "Award-Winning Computer Vision", "method": "Learning to Solve Hard Minimal Problems introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "Learning to Solve Hard Minimal Problems is an award-winning computer vision paper.", "motivation": "Learning to Solve Hard Minimal Problems targets a central problem recognized by the conference awards committee.", "implementation": "Learning to Solve Hard Minimal Problems introduces a method or system selected as award-winning work.", "application": "Learning to Solve Hard Minimal Problems is relevant to its conference track and downstream computer vision applications.", "problem": "Learning to Solve Hard Minimal Problems targets a central problem recognized by the conference awards committee.", "contribution": "Learning to Solve Hard Minimal Problems introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "CVPR 2022 Best Paper: Learning to Solve Hard Minimal Problems is an award-winning computer vision paper.", "whyFollow": "CVPR 2022 Best Paper，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "Learning to Solve Hard Minimal Problems is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "CVPR", "Best Paper", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "CVPR", "year": 2022, "award": "Best Paper", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "EPro-PnP: Generalized End-to-End Probabilistic Perspective-n-Points for Monocular Object Pose Estimation", "authors": "Hansheng Chen, Pichao Wang, Fan Wang, Tianyu Wang, Liwei Xiong, Hao Li", "venue": "CVPR", "year": 2022, "topic": "3D / Geometry / Reconstruction", "modality": "3D / Geometry / Reconstruction", "status": "CV 顶会 Best Paper 合集", "priority": "High", "task": "3D / Geometry / Reconstruction", "method": "EPro-PnP: Generalized End-to-End Probabilistic Perspective-n-Points for Monocular Object Pose Estimation introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "EPro-PnP: Generalized End-to-End Probabilistic Perspective-n-Points for Monocular Object Pose Estimation is an award-winning computer vision paper.", "motivation": "EPro-PnP: Generalized End-to-End Probabilistic Perspective-n-Points for Monocular Object Pose Estimation targets a central problem recognized by the conference awards committee.", "implementation": "EPro-PnP: Generalized End-to-End Probabilistic Perspective-n-Points for Monocular Object Pose Estimation introduces a method or system selected as award-winning work.", "application": "EPro-PnP: Generalized End-to-End Probabilistic Perspective-n-Points for Monocular Object Pose Estimation is relevant to its conference track and downstream computer vision applications.", "problem": "EPro-PnP: Generalized End-to-End Probabilistic Perspective-n-Points for Monocular Object Pose Estimation targets a central problem recognized by the conference awards committee.", "contribution": "EPro-PnP: Generalized End-to-End Probabilistic Perspective-n-Points for Monocular Object Pose Estimation introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "CVPR 2022 Best Student Paper: EPro-PnP: Generalized End-to-End Probabilistic Perspective-n-Points for Monocular Object Pose Estimation is an award-winning computer vision paper.", "whyFollow": "CVPR 2022 Best Student Paper，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "EPro-PnP: Generalized End-to-End Probabilistic Perspective-n-Points for Monocular Object Pose Estimation is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "CVPR", "Best Student", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "CVPR", "year": 2022, "award": "Best Student Paper", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "Dual-Shutter Optical Vibration Sensing", "authors": "Mark Sheinin, David Chan, Matthew O'Toole, Srinivasa Narasimhan", "venue": "CVPR", "year": 2022, "topic": "Computational Imaging", "modality": "Computational Imaging", "status": "CV 顶会 Best Paper 合集", "priority": "Medium", "task": "Computational Imaging", "method": "Dual-Shutter Optical Vibration Sensing introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "Dual-Shutter Optical Vibration Sensing is an award-winning computer vision paper.", "motivation": "Dual-Shutter Optical Vibration Sensing targets a central problem recognized by the conference awards committee.", "implementation": "Dual-Shutter Optical Vibration Sensing introduces a method or system selected as award-winning work.", "application": "Dual-Shutter Optical Vibration Sensing is relevant to its conference track and downstream computer vision applications.", "problem": "Dual-Shutter Optical Vibration Sensing targets a central problem recognized by the conference awards committee.", "contribution": "Dual-Shutter Optical Vibration Sensing introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "CVPR 2022 Honorable Mention: Dual-Shutter Optical Vibration Sensing is an award-winning computer vision paper.", "whyFollow": "CVPR 2022 Honorable Mention，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "Dual-Shutter Optical Vibration Sensing is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "CVPR", "Honorable Mention", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "CVPR", "year": 2022, "award": "Honorable Mention", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "Ref-NeRF: Structured View-Dependent Appearance for Neural Radiance Fields", "authors": "Dor Verbin, Peter Hedman, Ben Mildenhall, Todd Zickler, Jonathan T. Barron, Pratul P. Srinivasan", "venue": "CVPR", "year": 2022, "topic": "3D / Geometry / Reconstruction", "modality": "3D / Geometry / Reconstruction", "status": "CV 顶会 Best Paper 合集", "priority": "Medium", "task": "3D / Geometry / Reconstruction", "method": "Ref-NeRF: Structured View-Dependent Appearance for Neural Radiance Fields introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "Ref-NeRF: Structured View-Dependent Appearance for Neural Radiance Fields is an award-winning computer vision paper.", "motivation": "Ref-NeRF: Structured View-Dependent Appearance for Neural Radiance Fields targets a central problem recognized by the conference awards committee.", "implementation": "Ref-NeRF: Structured View-Dependent Appearance for Neural Radiance Fields introduces a method or system selected as award-winning work.", "application": "Ref-NeRF: Structured View-Dependent Appearance for Neural Radiance Fields is relevant to its conference track and downstream computer vision applications.", "problem": "Ref-NeRF: Structured View-Dependent Appearance for Neural Radiance Fields targets a central problem recognized by the conference awards committee.", "contribution": "Ref-NeRF: Structured View-Dependent Appearance for Neural Radiance Fields introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "CVPR 2022 Honorable Mention: Ref-NeRF: Structured View-Dependent Appearance for Neural Radiance Fields is an award-winning computer vision paper.", "whyFollow": "CVPR 2022 Honorable Mention，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "Ref-NeRF: Structured View-Dependent Appearance for Neural Radiance Fields is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "CVPR", "Honorable Mention", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "CVPR", "year": 2022, "award": "Honorable Mention", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "GIRAFFE: Representing Scenes as Compositional Generative Neural Feature Fields", "authors": "Michael Niemeyer, Andreas Geiger", "venue": "CVPR", "year": 2021, "topic": "Generative Vision", "modality": "Generative Vision", "status": "CV 顶会 Best Paper 合集", "priority": "High", "task": "Generative Vision", "method": "GIRAFFE: Representing Scenes as Compositional Generative Neural Feature Fields introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "GIRAFFE: Representing Scenes as Compositional Generative Neural Feature Fields is an award-winning computer vision paper.", "motivation": "GIRAFFE: Representing Scenes as Compositional Generative Neural Feature Fields targets a central problem recognized by the conference awards committee.", "implementation": "GIRAFFE: Representing Scenes as Compositional Generative Neural Feature Fields introduces a method or system selected as award-winning work.", "application": "GIRAFFE: Representing Scenes as Compositional Generative Neural Feature Fields is relevant to its conference track and downstream computer vision applications.", "problem": "GIRAFFE: Representing Scenes as Compositional Generative Neural Feature Fields targets a central problem recognized by the conference awards committee.", "contribution": "GIRAFFE: Representing Scenes as Compositional Generative Neural Feature Fields introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "CVPR 2021 Best Paper: GIRAFFE: Representing Scenes as Compositional Generative Neural Feature Fields is an award-winning computer vision paper.", "whyFollow": "CVPR 2021 Best Paper，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "GIRAFFE: Representing Scenes as Compositional Generative Neural Feature Fields is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "CVPR", "Best Paper", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "CVPR", "year": 2021, "award": "Best Paper", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "Task Programming: Learning Data Efficient Behavior Representations", "authors": "Jennifer J. Sun, Ann Kennedy, Eric Zhan, David J. Anderson, Yisong Yue, Pietro Perona", "venue": "CVPR", "year": 2021, "topic": "Vision Backbones", "modality": "Vision Backbones", "status": "CV 顶会 Best Paper 合集", "priority": "High", "task": "Vision Backbones", "method": "Task Programming: Learning Data Efficient Behavior Representations introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "Task Programming: Learning Data Efficient Behavior Representations is an award-winning computer vision paper.", "motivation": "Task Programming: Learning Data Efficient Behavior Representations targets a central problem recognized by the conference awards committee.", "implementation": "Task Programming: Learning Data Efficient Behavior Representations introduces a method or system selected as award-winning work.", "application": "Task Programming: Learning Data Efficient Behavior Representations is relevant to its conference track and downstream computer vision applications.", "problem": "Task Programming: Learning Data Efficient Behavior Representations targets a central problem recognized by the conference awards committee.", "contribution": "Task Programming: Learning Data Efficient Behavior Representations introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "CVPR 2021 Best Student Paper: Task Programming: Learning Data Efficient Behavior Representations is an award-winning computer vision paper.", "whyFollow": "CVPR 2021 Best Student Paper，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "Task Programming: Learning Data Efficient Behavior Representations is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "CVPR", "Best Student", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "CVPR", "year": 2021, "award": "Best Student Paper", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "Learning High Fidelity Depths of Dressed Humans by Watching Social Media Dance Videos", "authors": "Yasamin Jafarian, Hyun Soo Park", "venue": "CVPR", "year": 2021, "topic": "3D / Geometry / Reconstruction", "modality": "3D / Geometry / Reconstruction", "status": "CV 顶会 Best Paper 合集", "priority": "Medium", "task": "3D / Geometry / Reconstruction", "method": "Learning High Fidelity Depths of Dressed Humans by Watching Social Media Dance Videos introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "Learning High Fidelity Depths of Dressed Humans by Watching Social Media Dance Videos is an award-winning computer vision paper.", "motivation": "Learning High Fidelity Depths of Dressed Humans by Watching Social Media Dance Videos targets a central problem recognized by the conference awards committee.", "implementation": "Learning High Fidelity Depths of Dressed Humans by Watching Social Media Dance Videos introduces a method or system selected as award-winning work.", "application": "Learning High Fidelity Depths of Dressed Humans by Watching Social Media Dance Videos is relevant to its conference track and downstream computer vision applications.", "problem": "Learning High Fidelity Depths of Dressed Humans by Watching Social Media Dance Videos targets a central problem recognized by the conference awards committee.", "contribution": "Learning High Fidelity Depths of Dressed Humans by Watching Social Media Dance Videos introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "CVPR 2021 Honorable Mention: Learning High Fidelity Depths of Dressed Humans by Watching Social Media Dance Videos is an award-winning computer vision paper.", "whyFollow": "CVPR 2021 Honorable Mention，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "Learning High Fidelity Depths of Dressed Humans by Watching Social Media Dance Videos is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "CVPR", "Honorable Mention", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "CVPR", "year": 2021, "award": "Honorable Mention", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "Less is More: ClipBERT for Video-and-Language Learning via Sparse Sampling", "authors": "Jie Lei, Linjie Li, Luowei Zhou, Zhe Gan, Tamara L. Berg, Mohit Bansal, Jingjing Liu", "venue": "CVPR", "year": 2021, "topic": "Video / Embodied Perception", "modality": "Video / Embodied Perception", "status": "CV 顶会 Best Paper 合集", "priority": "Medium", "task": "Video / Embodied Perception", "method": "Less is More: ClipBERT for Video-and-Language Learning via Sparse Sampling introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "Less is More: ClipBERT for Video-and-Language Learning via Sparse Sampling is an award-winning computer vision paper.", "motivation": "Less is More: ClipBERT for Video-and-Language Learning via Sparse Sampling targets a central problem recognized by the conference awards committee.", "implementation": "Less is More: ClipBERT for Video-and-Language Learning via Sparse Sampling introduces a method or system selected as award-winning work.", "application": "Less is More: ClipBERT for Video-and-Language Learning via Sparse Sampling is relevant to its conference track and downstream computer vision applications.", "problem": "Less is More: ClipBERT for Video-and-Language Learning via Sparse Sampling targets a central problem recognized by the conference awards committee.", "contribution": "Less is More: ClipBERT for Video-and-Language Learning via Sparse Sampling introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "CVPR 2021 Honorable Mention: Less is More: ClipBERT for Video-and-Language Learning via Sparse Sampling is an award-winning computer vision paper.", "whyFollow": "CVPR 2021 Honorable Mention，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "Less is More: ClipBERT for Video-and-Language Learning via Sparse Sampling is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "CVPR", "Honorable Mention", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "CVPR", "year": 2021, "award": "Honorable Mention", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "Binary TTC: A Temporal Geofence for Autonomous Navigation", "authors": "Abhishek Badki, Orazio Gallo, Jan Kautz, Pradeep Sen", "venue": "CVPR", "year": 2021, "topic": "Video / Embodied Perception", "modality": "Video / Embodied Perception", "status": "CV 顶会 Best Paper 合集", "priority": "Medium", "task": "Video / Embodied Perception", "method": "Binary TTC: A Temporal Geofence for Autonomous Navigation introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "Binary TTC: A Temporal Geofence for Autonomous Navigation is an award-winning computer vision paper.", "motivation": "Binary TTC: A Temporal Geofence for Autonomous Navigation targets a central problem recognized by the conference awards committee.", "implementation": "Binary TTC: A Temporal Geofence for Autonomous Navigation introduces a method or system selected as award-winning work.", "application": "Binary TTC: A Temporal Geofence for Autonomous Navigation is relevant to its conference track and downstream computer vision applications.", "problem": "Binary TTC: A Temporal Geofence for Autonomous Navigation targets a central problem recognized by the conference awards committee.", "contribution": "Binary TTC: A Temporal Geofence for Autonomous Navigation introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "CVPR 2021 Honorable Mention: Binary TTC: A Temporal Geofence for Autonomous Navigation is an award-winning computer vision paper.", "whyFollow": "CVPR 2021 Honorable Mention，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "Binary TTC: A Temporal Geofence for Autonomous Navigation is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "CVPR", "Honorable Mention", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "CVPR", "year": 2021, "award": "Honorable Mention", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "Real-Time High-Resolution Background Matting", "authors": "Shanchuan Lin, Andrey Ryabtsev, Soumyadip Sengupta, Brian Curless, Steve Seitz, Ira Kemelmacher-Shlizerman", "venue": "CVPR", "year": 2021, "topic": "Segmentation", "modality": "Segmentation", "status": "CV 顶会 Best Paper 合集", "priority": "Medium", "task": "Segmentation", "method": "Real-Time High-Resolution Background Matting introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "Real-Time High-Resolution Background Matting is an award-winning computer vision paper.", "motivation": "Real-Time High-Resolution Background Matting targets a central problem recognized by the conference awards committee.", "implementation": "Real-Time High-Resolution Background Matting introduces a method or system selected as award-winning work.", "application": "Real-Time High-Resolution Background Matting is relevant to its conference track and downstream computer vision applications.", "problem": "Real-Time High-Resolution Background Matting targets a central problem recognized by the conference awards committee.", "contribution": "Real-Time High-Resolution Background Matting introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "CVPR 2021 Honorable Mention: Real-Time High-Resolution Background Matting is an award-winning computer vision paper.", "whyFollow": "CVPR 2021 Honorable Mention，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "Real-Time High-Resolution Background Matting is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "CVPR", "Honorable Mention", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "CVPR", "year": 2021, "award": "Honorable Mention", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "Unsupervised Learning of Probably Symmetric Deformable 3D Objects from Images in the Wild", "authors": "Shangzhe Wu, Christian Rupprecht, Andrea Vedaldi", "venue": "CVPR", "year": 2020, "topic": "3D / Geometry / Reconstruction", "modality": "3D / Geometry / Reconstruction", "status": "CV 顶会 Best Paper 合集", "priority": "High", "task": "3D / Geometry / Reconstruction", "method": "Unsupervised Learning of Probably Symmetric Deformable 3D Objects from Images in the Wild introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "Unsupervised Learning of Probably Symmetric Deformable 3D Objects from Images in the Wild is an award-winning computer vision paper.", "motivation": "Unsupervised Learning of Probably Symmetric Deformable 3D Objects from Images in the Wild targets a central problem recognized by the conference awards committee.", "implementation": "Unsupervised Learning of Probably Symmetric Deformable 3D Objects from Images in the Wild introduces a method or system selected as award-winning work.", "application": "Unsupervised Learning of Probably Symmetric Deformable 3D Objects from Images in the Wild is relevant to its conference track and downstream computer vision applications.", "problem": "Unsupervised Learning of Probably Symmetric Deformable 3D Objects from Images in the Wild targets a central problem recognized by the conference awards committee.", "contribution": "Unsupervised Learning of Probably Symmetric Deformable 3D Objects from Images in the Wild introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "CVPR 2020 Best Paper: Unsupervised Learning of Probably Symmetric Deformable 3D Objects from Images in the Wild is an award-winning computer vision paper.", "whyFollow": "CVPR 2020 Best Paper，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "Unsupervised Learning of Probably Symmetric Deformable 3D Objects from Images in the Wild is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "CVPR", "Best Paper", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "CVPR", "year": 2020, "award": "Best Paper", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "BSP-Net: Generating Compact Meshes via Binary Space Partitioning", "authors": "Zhiqin Chen, Andrea Tagliasacchi, Hao Zhang", "venue": "CVPR", "year": 2020, "topic": "3D / Geometry / Reconstruction", "modality": "3D / Geometry / Reconstruction", "status": "CV 顶会 Best Paper 合集", "priority": "High", "task": "3D / Geometry / Reconstruction", "method": "BSP-Net: Generating Compact Meshes via Binary Space Partitioning introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "BSP-Net: Generating Compact Meshes via Binary Space Partitioning is an award-winning computer vision paper.", "motivation": "BSP-Net: Generating Compact Meshes via Binary Space Partitioning targets a central problem recognized by the conference awards committee.", "implementation": "BSP-Net: Generating Compact Meshes via Binary Space Partitioning introduces a method or system selected as award-winning work.", "application": "BSP-Net: Generating Compact Meshes via Binary Space Partitioning is relevant to its conference track and downstream computer vision applications.", "problem": "BSP-Net: Generating Compact Meshes via Binary Space Partitioning targets a central problem recognized by the conference awards committee.", "contribution": "BSP-Net: Generating Compact Meshes via Binary Space Partitioning introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "CVPR 2020 Best Student Paper: BSP-Net: Generating Compact Meshes via Binary Space Partitioning is an award-winning computer vision paper.", "whyFollow": "CVPR 2020 Best Student Paper，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "BSP-Net: Generating Compact Meshes via Binary Space Partitioning is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "CVPR", "Best Student", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "CVPR", "year": 2020, "award": "Best Student Paper", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "DeepCap: Monocular Human Performance Capture Using Weak Supervision", "authors": "Marc Habermann, Weipeng Xu, Michael Zollhoefer, Gerard Pons-Moll, Christian Theobalt", "venue": "CVPR", "year": 2020, "topic": "Award-Winning Computer Vision", "modality": "Award-Winning Computer Vision", "status": "CV 顶会 Best Paper 合集", "priority": "Medium", "task": "Award-Winning Computer Vision", "method": "DeepCap: Monocular Human Performance Capture Using Weak Supervision introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "DeepCap: Monocular Human Performance Capture Using Weak Supervision is an award-winning computer vision paper.", "motivation": "DeepCap: Monocular Human Performance Capture Using Weak Supervision targets a central problem recognized by the conference awards committee.", "implementation": "DeepCap: Monocular Human Performance Capture Using Weak Supervision introduces a method or system selected as award-winning work.", "application": "DeepCap: Monocular Human Performance Capture Using Weak Supervision is relevant to its conference track and downstream computer vision applications.", "problem": "DeepCap: Monocular Human Performance Capture Using Weak Supervision targets a central problem recognized by the conference awards committee.", "contribution": "DeepCap: Monocular Human Performance Capture Using Weak Supervision introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "CVPR 2020 Honorable Mention: DeepCap: Monocular Human Performance Capture Using Weak Supervision is an award-winning computer vision paper.", "whyFollow": "CVPR 2020 Honorable Mention，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "DeepCap: Monocular Human Performance Capture Using Weak Supervision is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "CVPR", "Honorable Mention", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "CVPR", "year": 2020, "award": "Honorable Mention", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "A Theory of Fermat Paths for Non-Line-of-Sight Shape Reconstruction", "authors": "Shumian Xin, Sotiris Nousias, Kiriakos N. Kutulakos, Achuta Kadambi, Srinivasa G. Narasimhan, Ioannis Gkioulekas", "venue": "CVPR", "year": 2019, "topic": "3D / Geometry / Reconstruction", "modality": "3D / Geometry / Reconstruction", "status": "CV 顶会 Best Paper 合集", "priority": "High", "task": "3D / Geometry / Reconstruction", "method": "A Theory of Fermat Paths for Non-Line-of-Sight Shape Reconstruction introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "A Theory of Fermat Paths for Non-Line-of-Sight Shape Reconstruction is an award-winning computer vision paper.", "motivation": "A Theory of Fermat Paths for Non-Line-of-Sight Shape Reconstruction targets a central problem recognized by the conference awards committee.", "implementation": "A Theory of Fermat Paths for Non-Line-of-Sight Shape Reconstruction introduces a method or system selected as award-winning work.", "application": "A Theory of Fermat Paths for Non-Line-of-Sight Shape Reconstruction is relevant to its conference track and downstream computer vision applications.", "problem": "A Theory of Fermat Paths for Non-Line-of-Sight Shape Reconstruction targets a central problem recognized by the conference awards committee.", "contribution": "A Theory of Fermat Paths for Non-Line-of-Sight Shape Reconstruction introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "CVPR 2019 Best Paper: A Theory of Fermat Paths for Non-Line-of-Sight Shape Reconstruction is an award-winning computer vision paper.", "whyFollow": "CVPR 2019 Best Paper，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "A Theory of Fermat Paths for Non-Line-of-Sight Shape Reconstruction is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "CVPR", "Best Paper", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "CVPR", "year": 2019, "award": "Best Paper", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "Reinforced Cross-Modal Matching and Self-Supervised Imitation Learning for Vision-Language Navigation", "authors": "Xin Wang, Qiuyuan Huang, Asli Celikyilmaz, Jianfeng Gao, Dinghan Shen, Yuan-Fang Wang, William Yang Wang, Lei Zhang", "venue": "CVPR", "year": 2019, "topic": "Video / Embodied Perception", "modality": "Video / Embodied Perception", "status": "CV 顶会 Best Paper 合集", "priority": "High", "task": "Video / Embodied Perception", "method": "Reinforced Cross-Modal Matching and Self-Supervised Imitation Learning for Vision-Language Navigation introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "Reinforced Cross-Modal Matching and Self-Supervised Imitation Learning for Vision-Language Navigation is an award-winning computer vision paper.", "motivation": "Reinforced Cross-Modal Matching and Self-Supervised Imitation Learning for Vision-Language Navigation targets a central problem recognized by the conference awards committee.", "implementation": "Reinforced Cross-Modal Matching and Self-Supervised Imitation Learning for Vision-Language Navigation introduces a method or system selected as award-winning work.", "application": "Reinforced Cross-Modal Matching and Self-Supervised Imitation Learning for Vision-Language Navigation is relevant to its conference track and downstream computer vision applications.", "problem": "Reinforced Cross-Modal Matching and Self-Supervised Imitation Learning for Vision-Language Navigation targets a central problem recognized by the conference awards committee.", "contribution": "Reinforced Cross-Modal Matching and Self-Supervised Imitation Learning for Vision-Language Navigation introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "CVPR 2019 Best Student Paper: Reinforced Cross-Modal Matching and Self-Supervised Imitation Learning for Vision-Language Navigation is an award-winning computer vision paper.", "whyFollow": "CVPR 2019 Best Student Paper，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "Reinforced Cross-Modal Matching and Self-Supervised Imitation Learning for Vision-Language Navigation is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "CVPR", "Best Student", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "CVPR", "year": 2019, "award": "Best Student Paper", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "Learning the Depths of Moving People by Watching Frozen People", "authors": "Zhengqi Li, Tali Dekel, Forrester Cole, Richard Tucker, Ce Liu, William T. Freeman, Noah Snavely", "venue": "CVPR", "year": 2019, "topic": "3D / Geometry / Reconstruction", "modality": "3D / Geometry / Reconstruction", "status": "CV 顶会 Best Paper 合集", "priority": "Medium", "task": "3D / Geometry / Reconstruction", "method": "Learning the Depths of Moving People by Watching Frozen People introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "Learning the Depths of Moving People by Watching Frozen People is an award-winning computer vision paper.", "motivation": "Learning the Depths of Moving People by Watching Frozen People targets a central problem recognized by the conference awards committee.", "implementation": "Learning the Depths of Moving People by Watching Frozen People introduces a method or system selected as award-winning work.", "application": "Learning the Depths of Moving People by Watching Frozen People is relevant to its conference track and downstream computer vision applications.", "problem": "Learning the Depths of Moving People by Watching Frozen People targets a central problem recognized by the conference awards committee.", "contribution": "Learning the Depths of Moving People by Watching Frozen People introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "CVPR 2019 Honorable Mention: Learning the Depths of Moving People by Watching Frozen People is an award-winning computer vision paper.", "whyFollow": "CVPR 2019 Honorable Mention，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "Learning the Depths of Moving People by Watching Frozen People is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "CVPR", "Honorable Mention", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "CVPR", "year": 2019, "award": "Honorable Mention", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "Taskonomy: Disentangling Task Transfer Learning", "authors": "Amir R. Zamir, Alexander Sax, William Shen, Leonidas J. Guibas, Jitendra Malik, Silvio Savarese", "venue": "CVPR", "year": 2018, "topic": "Award-Winning Computer Vision", "modality": "Award-Winning Computer Vision", "status": "CV 顶会 Best Paper 合集", "priority": "High", "task": "Award-Winning Computer Vision", "method": "Taskonomy: Disentangling Task Transfer Learning introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "Taskonomy: Disentangling Task Transfer Learning is an award-winning computer vision paper.", "motivation": "Taskonomy: Disentangling Task Transfer Learning targets a central problem recognized by the conference awards committee.", "implementation": "Taskonomy: Disentangling Task Transfer Learning introduces a method or system selected as award-winning work.", "application": "Taskonomy: Disentangling Task Transfer Learning is relevant to its conference track and downstream computer vision applications.", "problem": "Taskonomy: Disentangling Task Transfer Learning targets a central problem recognized by the conference awards committee.", "contribution": "Taskonomy: Disentangling Task Transfer Learning introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "CVPR 2018 Best Paper: Taskonomy: Disentangling Task Transfer Learning is an award-winning computer vision paper.", "whyFollow": "CVPR 2018 Best Paper，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "Taskonomy: Disentangling Task Transfer Learning is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "CVPR", "Best Paper", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "CVPR", "year": 2018, "award": "Best Paper", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "Total Capture: A 3D Deformation Model for Tracking Faces, Hands, and Bodies", "authors": "Hanbyul Joo, Tomas Simon, Yaser Sheikh", "venue": "CVPR", "year": 2018, "topic": "3D / Geometry / Reconstruction", "modality": "3D / Geometry / Reconstruction", "status": "CV 顶会 Best Paper 合集", "priority": "High", "task": "3D / Geometry / Reconstruction", "method": "Total Capture: A 3D Deformation Model for Tracking Faces, Hands, and Bodies introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "Total Capture: A 3D Deformation Model for Tracking Faces, Hands, and Bodies is an award-winning computer vision paper.", "motivation": "Total Capture: A 3D Deformation Model for Tracking Faces, Hands, and Bodies targets a central problem recognized by the conference awards committee.", "implementation": "Total Capture: A 3D Deformation Model for Tracking Faces, Hands, and Bodies introduces a method or system selected as award-winning work.", "application": "Total Capture: A 3D Deformation Model for Tracking Faces, Hands, and Bodies is relevant to its conference track and downstream computer vision applications.", "problem": "Total Capture: A 3D Deformation Model for Tracking Faces, Hands, and Bodies targets a central problem recognized by the conference awards committee.", "contribution": "Total Capture: A 3D Deformation Model for Tracking Faces, Hands, and Bodies introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "CVPR 2018 Best Student Paper: Total Capture: A 3D Deformation Model for Tracking Faces, Hands, and Bodies is an award-winning computer vision paper.", "whyFollow": "CVPR 2018 Best Student Paper，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "Total Capture: A 3D Deformation Model for Tracking Faces, Hands, and Bodies is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "CVPR", "Best Student", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "CVPR", "year": 2018, "award": "Best Student Paper", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "Deep Learning of Graph Matching", "authors": "Andrei Zanfir, Cristian Sminchisescu", "venue": "CVPR", "year": 2018, "topic": "Award-Winning Computer Vision", "modality": "Award-Winning Computer Vision", "status": "CV 顶会 Best Paper 合集", "priority": "Medium", "task": "Award-Winning Computer Vision", "method": "Deep Learning of Graph Matching introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "Deep Learning of Graph Matching is an award-winning computer vision paper.", "motivation": "Deep Learning of Graph Matching targets a central problem recognized by the conference awards committee.", "implementation": "Deep Learning of Graph Matching introduces a method or system selected as award-winning work.", "application": "Deep Learning of Graph Matching is relevant to its conference track and downstream computer vision applications.", "problem": "Deep Learning of Graph Matching targets a central problem recognized by the conference awards committee.", "contribution": "Deep Learning of Graph Matching introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "CVPR 2018 Honorable Mention: Deep Learning of Graph Matching is an award-winning computer vision paper.", "whyFollow": "CVPR 2018 Honorable Mention，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "Deep Learning of Graph Matching is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "CVPR", "Honorable Mention", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "CVPR", "year": 2018, "award": "Honorable Mention", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "SPLATNet: Sparse Lattice Networks for Point Cloud Processing", "authors": "Hang Su, Varun Jampani, Deqing Sun, Subhransu Maji, Evangelos Kalogerakis, Ming-Hsuan Yang, Jan Kautz", "venue": "CVPR", "year": 2018, "topic": "3D / Geometry / Reconstruction", "modality": "3D / Geometry / Reconstruction", "status": "CV 顶会 Best Paper 合集", "priority": "Medium", "task": "3D / Geometry / Reconstruction", "method": "SPLATNet: Sparse Lattice Networks for Point Cloud Processing introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "SPLATNet: Sparse Lattice Networks for Point Cloud Processing is an award-winning computer vision paper.", "motivation": "SPLATNet: Sparse Lattice Networks for Point Cloud Processing targets a central problem recognized by the conference awards committee.", "implementation": "SPLATNet: Sparse Lattice Networks for Point Cloud Processing introduces a method or system selected as award-winning work.", "application": "SPLATNet: Sparse Lattice Networks for Point Cloud Processing is relevant to its conference track and downstream computer vision applications.", "problem": "SPLATNet: Sparse Lattice Networks for Point Cloud Processing targets a central problem recognized by the conference awards committee.", "contribution": "SPLATNet: Sparse Lattice Networks for Point Cloud Processing introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "CVPR 2018 Honorable Mention: SPLATNet: Sparse Lattice Networks for Point Cloud Processing is an award-winning computer vision paper.", "whyFollow": "CVPR 2018 Honorable Mention，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "SPLATNet: Sparse Lattice Networks for Point Cloud Processing is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "CVPR", "Honorable Mention", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "CVPR", "year": 2018, "award": "Honorable Mention", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "CodeSLAM: Learning a Compact, Optimisable Representation for Dense Visual SLAM", "authors": "Michael Bloesch, Jan Czarnowski, Ronald Clark, Stefan Leutenegger, Andrew J. Davison", "venue": "CVPR", "year": 2018, "topic": "3D / Geometry / Reconstruction", "modality": "3D / Geometry / Reconstruction", "status": "CV 顶会 Best Paper 合集", "priority": "Medium", "task": "3D / Geometry / Reconstruction", "method": "CodeSLAM: Learning a Compact, Optimisable Representation for Dense Visual SLAM introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "CodeSLAM: Learning a Compact, Optimisable Representation for Dense Visual SLAM is an award-winning computer vision paper.", "motivation": "CodeSLAM: Learning a Compact, Optimisable Representation for Dense Visual SLAM targets a central problem recognized by the conference awards committee.", "implementation": "CodeSLAM: Learning a Compact, Optimisable Representation for Dense Visual SLAM introduces a method or system selected as award-winning work.", "application": "CodeSLAM: Learning a Compact, Optimisable Representation for Dense Visual SLAM is relevant to its conference track and downstream computer vision applications.", "problem": "CodeSLAM: Learning a Compact, Optimisable Representation for Dense Visual SLAM targets a central problem recognized by the conference awards committee.", "contribution": "CodeSLAM: Learning a Compact, Optimisable Representation for Dense Visual SLAM introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "CVPR 2018 Honorable Mention: CodeSLAM: Learning a Compact, Optimisable Representation for Dense Visual SLAM is an award-winning computer vision paper.", "whyFollow": "CVPR 2018 Honorable Mention，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "CodeSLAM: Learning a Compact, Optimisable Representation for Dense Visual SLAM is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "CVPR", "Honorable Mention", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "CVPR", "year": 2018, "award": "Honorable Mention", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "Efficient Optimization for Rank-Based Loss Functions", "authors": "Pranay Mohapatra, Michael Rolinek, C. V. Jawahar, Vladimir Kolmogorov, M. Pawan Kumar", "venue": "CVPR", "year": 2018, "topic": "Award-Winning Computer Vision", "modality": "Award-Winning Computer Vision", "status": "CV 顶会 Best Paper 合集", "priority": "Medium", "task": "Award-Winning Computer Vision", "method": "Efficient Optimization for Rank-Based Loss Functions introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "Efficient Optimization for Rank-Based Loss Functions is an award-winning computer vision paper.", "motivation": "Efficient Optimization for Rank-Based Loss Functions targets a central problem recognized by the conference awards committee.", "implementation": "Efficient Optimization for Rank-Based Loss Functions introduces a method or system selected as award-winning work.", "application": "Efficient Optimization for Rank-Based Loss Functions is relevant to its conference track and downstream computer vision applications.", "problem": "Efficient Optimization for Rank-Based Loss Functions targets a central problem recognized by the conference awards committee.", "contribution": "Efficient Optimization for Rank-Based Loss Functions introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "CVPR 2018 Honorable Mention: Efficient Optimization for Rank-Based Loss Functions is an award-winning computer vision paper.", "whyFollow": "CVPR 2018 Honorable Mention，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "Efficient Optimization for Rank-Based Loss Functions is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "CVPR", "Honorable Mention", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "CVPR", "year": 2018, "award": "Honorable Mention", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "Learning from Simulated and Unsupervised Images through Adversarial Training", "authors": "Ashish Shrivastava, Tomas Pfister, Oncel Tuzel, Joshua Susskind, Wenda Wang, Russ Webb", "venue": "CVPR", "year": 2017, "topic": "Award-Winning Computer Vision", "modality": "Award-Winning Computer Vision", "status": "CV 顶会 Best Paper 合集", "priority": "High", "task": "Award-Winning Computer Vision", "method": "Learning from Simulated and Unsupervised Images through Adversarial Training introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "Learning from Simulated and Unsupervised Images through Adversarial Training is an award-winning computer vision paper.", "motivation": "Learning from Simulated and Unsupervised Images through Adversarial Training targets a central problem recognized by the conference awards committee.", "implementation": "Learning from Simulated and Unsupervised Images through Adversarial Training introduces a method or system selected as award-winning work.", "application": "Learning from Simulated and Unsupervised Images through Adversarial Training is relevant to its conference track and downstream computer vision applications.", "problem": "Learning from Simulated and Unsupervised Images through Adversarial Training targets a central problem recognized by the conference awards committee.", "contribution": "Learning from Simulated and Unsupervised Images through Adversarial Training introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "CVPR 2017 Best Paper: Learning from Simulated and Unsupervised Images through Adversarial Training is an award-winning computer vision paper.", "whyFollow": "CVPR 2017 Best Paper，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "Learning from Simulated and Unsupervised Images through Adversarial Training is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "CVPR", "Best Paper", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "CVPR", "year": 2017, "award": "Best Paper", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "Computational Imaging on the Electric Grid", "authors": "Mark Sheinin, Yoav Y. Schechner, Kiriakos N. Kutulakos", "venue": "CVPR", "year": 2017, "topic": "Computational Imaging", "modality": "Computational Imaging", "status": "CV 顶会 Best Paper 合集", "priority": "High", "task": "Computational Imaging", "method": "Computational Imaging on the Electric Grid introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "Computational Imaging on the Electric Grid is an award-winning computer vision paper.", "motivation": "Computational Imaging on the Electric Grid targets a central problem recognized by the conference awards committee.", "implementation": "Computational Imaging on the Electric Grid introduces a method or system selected as award-winning work.", "application": "Computational Imaging on the Electric Grid is relevant to its conference track and downstream computer vision applications.", "problem": "Computational Imaging on the Electric Grid targets a central problem recognized by the conference awards committee.", "contribution": "Computational Imaging on the Electric Grid introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "CVPR 2017 Best Student Paper: Computational Imaging on the Electric Grid is an award-winning computer vision paper.", "whyFollow": "CVPR 2017 Best Student Paper，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "Computational Imaging on the Electric Grid is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "CVPR", "Best Student", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "CVPR", "year": 2017, "award": "Best Student Paper", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "Annotating Object Instances with a Polygon-RNN", "authors": "Lluis Castrejon, Kaustav Kundu, Raquel Urtasun, Sanja Fidler", "venue": "CVPR", "year": 2017, "topic": "Object Detection", "modality": "Object Detection", "status": "CV 顶会 Best Paper 合集", "priority": "Medium", "task": "Object Detection", "method": "Annotating Object Instances with a Polygon-RNN introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "Annotating Object Instances with a Polygon-RNN is an award-winning computer vision paper.", "motivation": "Annotating Object Instances with a Polygon-RNN targets a central problem recognized by the conference awards committee.", "implementation": "Annotating Object Instances with a Polygon-RNN introduces a method or system selected as award-winning work.", "application": "Annotating Object Instances with a Polygon-RNN is relevant to its conference track and downstream computer vision applications.", "problem": "Annotating Object Instances with a Polygon-RNN targets a central problem recognized by the conference awards committee.", "contribution": "Annotating Object Instances with a Polygon-RNN introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "CVPR 2017 Honorable Mention: Annotating Object Instances with a Polygon-RNN is an award-winning computer vision paper.", "whyFollow": "CVPR 2017 Honorable Mention，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "Annotating Object Instances with a Polygon-RNN is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "CVPR", "Honorable Mention", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "CVPR", "year": 2017, "award": "Honorable Mention", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "Deep Residual Learning for Image Recognition", "authors": "Kaiming He, Xiangyu Zhang, Shaoqing Ren, Jian Sun", "venue": "CVPR", "year": 2016, "topic": "Award-Winning Computer Vision", "modality": "Award-Winning Computer Vision", "status": "CV 顶会 Best Paper 合集", "priority": "High", "task": "Award-Winning Computer Vision", "method": "Deep Residual Learning for Image Recognition introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "Deep Residual Learning for Image Recognition is an award-winning computer vision paper.", "motivation": "Deep Residual Learning for Image Recognition targets a central problem recognized by the conference awards committee.", "implementation": "Deep Residual Learning for Image Recognition introduces a method or system selected as award-winning work.", "application": "Deep Residual Learning for Image Recognition is relevant to its conference track and downstream computer vision applications.", "problem": "Deep Residual Learning for Image Recognition targets a central problem recognized by the conference awards committee.", "contribution": "Deep Residual Learning for Image Recognition introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "CVPR 2016 Best Paper: Deep Residual Learning for Image Recognition is an award-winning computer vision paper.", "whyFollow": "CVPR 2016 Best Paper，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "Deep Residual Learning for Image Recognition is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "CVPR", "Best Paper", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "CVPR", "year": 2016, "award": "Best Paper", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "Structural-RNN: Deep Learning on Spatio-Temporal Graphs", "authors": "Ashesh Jain, Amir R. Zamir, Silvio Savarese, Ashutosh Saxena", "venue": "CVPR", "year": 2016, "topic": "Award-Winning Computer Vision", "modality": "Award-Winning Computer Vision", "status": "CV 顶会 Best Paper 合集", "priority": "High", "task": "Award-Winning Computer Vision", "method": "Structural-RNN: Deep Learning on Spatio-Temporal Graphs introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "Structural-RNN: Deep Learning on Spatio-Temporal Graphs is an award-winning computer vision paper.", "motivation": "Structural-RNN: Deep Learning on Spatio-Temporal Graphs targets a central problem recognized by the conference awards committee.", "implementation": "Structural-RNN: Deep Learning on Spatio-Temporal Graphs introduces a method or system selected as award-winning work.", "application": "Structural-RNN: Deep Learning on Spatio-Temporal Graphs is relevant to its conference track and downstream computer vision applications.", "problem": "Structural-RNN: Deep Learning on Spatio-Temporal Graphs targets a central problem recognized by the conference awards committee.", "contribution": "Structural-RNN: Deep Learning on Spatio-Temporal Graphs introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "CVPR 2016 Best Student Paper: Structural-RNN: Deep Learning on Spatio-Temporal Graphs is an award-winning computer vision paper.", "whyFollow": "CVPR 2016 Best Student Paper，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "Structural-RNN: Deep Learning on Spatio-Temporal Graphs is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "CVPR", "Best Student", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "CVPR", "year": 2016, "award": "Best Student Paper", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "Sublabel-Accurate Relaxation of Nonconvex Energies", "authors": "Thomas Moellenhoff, Evgeny Laude, Michael Moeller, Jan Lellmann, Daniel Cremers", "venue": "CVPR", "year": 2016, "topic": "Award-Winning Computer Vision", "modality": "Award-Winning Computer Vision", "status": "CV 顶会 Best Paper 合集", "priority": "Medium", "task": "Award-Winning Computer Vision", "method": "Sublabel-Accurate Relaxation of Nonconvex Energies introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "Sublabel-Accurate Relaxation of Nonconvex Energies is an award-winning computer vision paper.", "motivation": "Sublabel-Accurate Relaxation of Nonconvex Energies targets a central problem recognized by the conference awards committee.", "implementation": "Sublabel-Accurate Relaxation of Nonconvex Energies introduces a method or system selected as award-winning work.", "application": "Sublabel-Accurate Relaxation of Nonconvex Energies is relevant to its conference track and downstream computer vision applications.", "problem": "Sublabel-Accurate Relaxation of Nonconvex Energies targets a central problem recognized by the conference awards committee.", "contribution": "Sublabel-Accurate Relaxation of Nonconvex Energies introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "CVPR 2016 Honorable Mention: Sublabel-Accurate Relaxation of Nonconvex Energies is an award-winning computer vision paper.", "whyFollow": "CVPR 2016 Honorable Mention，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "Sublabel-Accurate Relaxation of Nonconvex Energies is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "CVPR", "Honorable Mention", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "CVPR", "year": 2016, "award": "Honorable Mention", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "Generating Physically Stable and Buildable Brick Structures from Text", "authors": "Aaron Pun, Kangxue Deng, Ruoshi Liu, Deva Ramanan, C. Karen Liu, Jun-Yan Zhu", "venue": "ICCV", "year": 2025, "topic": "3D / Geometry / Reconstruction", "modality": "3D / Geometry / Reconstruction", "status": "CV 顶会 Best Paper 合集", "priority": "High", "task": "3D / Geometry / Reconstruction", "method": "Generating Physically Stable and Buildable Brick Structures from Text introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "Generating Physically Stable and Buildable Brick Structures from Text is an award-winning computer vision paper.", "motivation": "Generating Physically Stable and Buildable Brick Structures from Text targets a central problem recognized by the conference awards committee.", "implementation": "Generating Physically Stable and Buildable Brick Structures from Text introduces a method or system selected as award-winning work.", "application": "Generating Physically Stable and Buildable Brick Structures from Text is relevant to its conference track and downstream computer vision applications.", "problem": "Generating Physically Stable and Buildable Brick Structures from Text targets a central problem recognized by the conference awards committee.", "contribution": "Generating Physically Stable and Buildable Brick Structures from Text introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "ICCV 2025 Best Paper: Generating Physically Stable and Buildable Brick Structures from Text is an award-winning computer vision paper.", "whyFollow": "ICCV 2025 Best Paper，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "Generating Physically Stable and Buildable Brick Structures from Text is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "ICCV", "Best Paper", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "ICCV", "year": 2025, "award": "Best Paper", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "FlowEdit: Inversion-Free Text-Based Editing Using Pre-Trained Flow Models", "authors": "Vladimir Kulikov, Matan Kleiner, Inbar Huberman-Spiegelglas, Tomer Michaeli", "venue": "ICCV", "year": 2025, "topic": "Award-Winning Computer Vision", "modality": "Award-Winning Computer Vision", "status": "CV 顶会 Best Paper 合集", "priority": "High", "task": "Award-Winning Computer Vision", "method": "FlowEdit: Inversion-Free Text-Based Editing Using Pre-Trained Flow Models introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "FlowEdit: Inversion-Free Text-Based Editing Using Pre-Trained Flow Models is an award-winning computer vision paper.", "motivation": "FlowEdit: Inversion-Free Text-Based Editing Using Pre-Trained Flow Models targets a central problem recognized by the conference awards committee.", "implementation": "FlowEdit: Inversion-Free Text-Based Editing Using Pre-Trained Flow Models introduces a method or system selected as award-winning work.", "application": "FlowEdit: Inversion-Free Text-Based Editing Using Pre-Trained Flow Models is relevant to its conference track and downstream computer vision applications.", "problem": "FlowEdit: Inversion-Free Text-Based Editing Using Pre-Trained Flow Models targets a central problem recognized by the conference awards committee.", "contribution": "FlowEdit: Inversion-Free Text-Based Editing Using Pre-Trained Flow Models introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "ICCV 2025 Best Student Paper: FlowEdit: Inversion-Free Text-Based Editing Using Pre-Trained Flow Models is an award-winning computer vision paper.", "whyFollow": "ICCV 2025 Best Student Paper，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "FlowEdit: Inversion-Free Text-Based Editing Using Pre-Trained Flow Models is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "ICCV", "Best Student", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "ICCV", "year": 2025, "award": "Best Student Paper", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "Spatially-Varying Autofocus", "authors": "Yuqian Qin, Aswin C. Sankaranarayanan, Matthew O'Toole", "venue": "ICCV", "year": 2025, "topic": "Computational Imaging", "modality": "Computational Imaging", "status": "CV 顶会 Best Paper 合集", "priority": "Medium", "task": "Computational Imaging", "method": "Spatially-Varying Autofocus introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "Spatially-Varying Autofocus is an award-winning computer vision paper.", "motivation": "Spatially-Varying Autofocus targets a central problem recognized by the conference awards committee.", "implementation": "Spatially-Varying Autofocus introduces a method or system selected as award-winning work.", "application": "Spatially-Varying Autofocus is relevant to its conference track and downstream computer vision applications.", "problem": "Spatially-Varying Autofocus targets a central problem recognized by the conference awards committee.", "contribution": "Spatially-Varying Autofocus introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "ICCV 2025 Honorable Mention: Spatially-Varying Autofocus is an award-winning computer vision paper.", "whyFollow": "ICCV 2025 Honorable Mention，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "Spatially-Varying Autofocus is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "ICCV", "Honorable Mention", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "ICCV", "year": 2025, "award": "Honorable Mention", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "RayZer: A Self-supervised Large View Synthesis Model", "authors": "Hanwen Jiang, Hao Tan, Peng Wang, Haimin Jin, Yao Zhao, Sai Bi, Kai Zhang, Fujun Luan, Kalyan Sunkavalli, Qixing Huang, Georgios Pavlakos", "venue": "ICCV", "year": 2025, "topic": "3D / Geometry / Reconstruction", "modality": "3D / Geometry / Reconstruction", "status": "CV 顶会 Best Paper 合集", "priority": "Medium", "task": "3D / Geometry / Reconstruction", "method": "RayZer: A Self-supervised Large View Synthesis Model introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "RayZer: A Self-supervised Large View Synthesis Model is an award-winning computer vision paper.", "motivation": "RayZer: A Self-supervised Large View Synthesis Model targets a central problem recognized by the conference awards committee.", "implementation": "RayZer: A Self-supervised Large View Synthesis Model introduces a method or system selected as award-winning work.", "application": "RayZer: A Self-supervised Large View Synthesis Model is relevant to its conference track and downstream computer vision applications.", "problem": "RayZer: A Self-supervised Large View Synthesis Model targets a central problem recognized by the conference awards committee.", "contribution": "RayZer: A Self-supervised Large View Synthesis Model introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "ICCV 2025 Honorable Mention: RayZer: A Self-supervised Large View Synthesis Model is an award-winning computer vision paper.", "whyFollow": "ICCV 2025 Honorable Mention，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "RayZer: A Self-supervised Large View Synthesis Model is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "ICCV", "Honorable Mention", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "ICCV", "year": 2025, "award": "Honorable Mention", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "Passive Ultra-Wideband Single-Photon Imaging", "authors": "Muyang Wei, Sotiris Nousias, Rahul Gulve, David B. Lindell, Kiriakos N. Kutulakos", "venue": "ICCV", "year": 2023, "topic": "Computational Imaging", "modality": "Computational Imaging", "status": "CV 顶会 Best Paper 合集", "priority": "High", "task": "Computational Imaging", "method": "Passive Ultra-Wideband Single-Photon Imaging introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "Passive Ultra-Wideband Single-Photon Imaging is an award-winning computer vision paper.", "motivation": "Passive Ultra-Wideband Single-Photon Imaging targets a central problem recognized by the conference awards committee.", "implementation": "Passive Ultra-Wideband Single-Photon Imaging introduces a method or system selected as award-winning work.", "application": "Passive Ultra-Wideband Single-Photon Imaging is relevant to its conference track and downstream computer vision applications.", "problem": "Passive Ultra-Wideband Single-Photon Imaging targets a central problem recognized by the conference awards committee.", "contribution": "Passive Ultra-Wideband Single-Photon Imaging introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "ICCV 2023 Best Paper: Passive Ultra-Wideband Single-Photon Imaging is an award-winning computer vision paper.", "whyFollow": "ICCV 2023 Best Paper，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "Passive Ultra-Wideband Single-Photon Imaging is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "ICCV", "Best Paper", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "ICCV", "year": 2023, "award": "Best Paper", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "Tracking Everything Everywhere All at Once", "authors": "Qianqian Wang, Yen-Yu Chang, Ruojin Cai, Zhengqi Li, Bharath Hariharan, Aleksander Holynski, Noah Snavely", "venue": "ICCV", "year": 2023, "topic": "Award-Winning Computer Vision", "modality": "Award-Winning Computer Vision", "status": "CV 顶会 Best Paper 合集", "priority": "High", "task": "Award-Winning Computer Vision", "method": "Tracking Everything Everywhere All at Once introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "Tracking Everything Everywhere All at Once is an award-winning computer vision paper.", "motivation": "Tracking Everything Everywhere All at Once targets a central problem recognized by the conference awards committee.", "implementation": "Tracking Everything Everywhere All at Once introduces a method or system selected as award-winning work.", "application": "Tracking Everything Everywhere All at Once is relevant to its conference track and downstream computer vision applications.", "problem": "Tracking Everything Everywhere All at Once targets a central problem recognized by the conference awards committee.", "contribution": "Tracking Everything Everywhere All at Once introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "ICCV 2023 Best Student Paper: Tracking Everything Everywhere All at Once is an award-winning computer vision paper.", "whyFollow": "ICCV 2023 Best Student Paper，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "Tracking Everything Everywhere All at Once is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "ICCV", "Best Student", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "ICCV", "year": 2023, "award": "Best Student Paper", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "Pixel-Perfect Structure-from-Motion with Featuremetric Refinement", "authors": "Philipp Lindenberger, Paul-Edouard Sarlin, Viktor Larsson, Marc Pollefeys", "venue": "ICCV", "year": 2021, "topic": "3D / Geometry / Reconstruction", "modality": "3D / Geometry / Reconstruction", "status": "CV 顶会 Best Paper 合集", "priority": "High", "task": "3D / Geometry / Reconstruction", "method": "Pixel-Perfect Structure-from-Motion with Featuremetric Refinement introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "Pixel-Perfect Structure-from-Motion with Featuremetric Refinement is an award-winning computer vision paper.", "motivation": "Pixel-Perfect Structure-from-Motion with Featuremetric Refinement targets a central problem recognized by the conference awards committee.", "implementation": "Pixel-Perfect Structure-from-Motion with Featuremetric Refinement introduces a method or system selected as award-winning work.", "application": "Pixel-Perfect Structure-from-Motion with Featuremetric Refinement is relevant to its conference track and downstream computer vision applications.", "problem": "Pixel-Perfect Structure-from-Motion with Featuremetric Refinement targets a central problem recognized by the conference awards committee.", "contribution": "Pixel-Perfect Structure-from-Motion with Featuremetric Refinement introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "ICCV 2021 Best Student Paper: Pixel-Perfect Structure-from-Motion with Featuremetric Refinement is an award-winning computer vision paper.", "whyFollow": "ICCV 2021 Best Student Paper，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "Pixel-Perfect Structure-from-Motion with Featuremetric Refinement is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "ICCV", "Best Student", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "ICCV", "year": 2021, "award": "Best Student Paper", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "Mip-NeRF: A Multiscale Representation for Anti-Aliasing Neural Radiance Fields", "authors": "Jonathan T. Barron, Ben Mildenhall, Matthew Tancik, Peter Hedman, Ricardo Martin-Brualla, Pratul P. Srinivasan", "venue": "ICCV", "year": 2021, "topic": "3D / Geometry / Reconstruction", "modality": "3D / Geometry / Reconstruction", "status": "CV 顶会 Best Paper 合集", "priority": "Medium", "task": "3D / Geometry / Reconstruction", "method": "Mip-NeRF: A Multiscale Representation for Anti-Aliasing Neural Radiance Fields introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "Mip-NeRF: A Multiscale Representation for Anti-Aliasing Neural Radiance Fields is an award-winning computer vision paper.", "motivation": "Mip-NeRF: A Multiscale Representation for Anti-Aliasing Neural Radiance Fields targets a central problem recognized by the conference awards committee.", "implementation": "Mip-NeRF: A Multiscale Representation for Anti-Aliasing Neural Radiance Fields introduces a method or system selected as award-winning work.", "application": "Mip-NeRF: A Multiscale Representation for Anti-Aliasing Neural Radiance Fields is relevant to its conference track and downstream computer vision applications.", "problem": "Mip-NeRF: A Multiscale Representation for Anti-Aliasing Neural Radiance Fields targets a central problem recognized by the conference awards committee.", "contribution": "Mip-NeRF: A Multiscale Representation for Anti-Aliasing Neural Radiance Fields introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "ICCV 2021 Honorable Mention: Mip-NeRF: A Multiscale Representation for Anti-Aliasing Neural Radiance Fields is an award-winning computer vision paper.", "whyFollow": "ICCV 2021 Honorable Mention，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "Mip-NeRF: A Multiscale Representation for Anti-Aliasing Neural Radiance Fields is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "ICCV", "Honorable Mention", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "ICCV", "year": 2021, "award": "Honorable Mention", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "OpenGAN: Open-Set Recognition via Open Data Generation", "authors": "Shu Kong, Deva Ramanan", "venue": "ICCV", "year": 2021, "topic": "Generative Vision", "modality": "Generative Vision", "status": "CV 顶会 Best Paper 合集", "priority": "Medium", "task": "Generative Vision", "method": "OpenGAN: Open-Set Recognition via Open Data Generation introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "OpenGAN: Open-Set Recognition via Open Data Generation is an award-winning computer vision paper.", "motivation": "OpenGAN: Open-Set Recognition via Open Data Generation targets a central problem recognized by the conference awards committee.", "implementation": "OpenGAN: Open-Set Recognition via Open Data Generation introduces a method or system selected as award-winning work.", "application": "OpenGAN: Open-Set Recognition via Open Data Generation is relevant to its conference track and downstream computer vision applications.", "problem": "OpenGAN: Open-Set Recognition via Open Data Generation targets a central problem recognized by the conference awards committee.", "contribution": "OpenGAN: Open-Set Recognition via Open Data Generation introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "ICCV 2021 Honorable Mention: OpenGAN: Open-Set Recognition via Open Data Generation is an award-winning computer vision paper.", "whyFollow": "ICCV 2021 Honorable Mention，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "OpenGAN: Open-Set Recognition via Open Data Generation is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "ICCV", "Honorable Mention", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "ICCV", "year": 2021, "award": "Honorable Mention", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "Viewing Graph Solvability via Cycle Consistency", "authors": "Federica Arrigoni, Andrea Fusiello, Elisa Ricci, Tomas Pajdla", "venue": "ICCV", "year": 2021, "topic": "3D / Geometry / Reconstruction", "modality": "3D / Geometry / Reconstruction", "status": "CV 顶会 Best Paper 合集", "priority": "Medium", "task": "3D / Geometry / Reconstruction", "method": "Viewing Graph Solvability via Cycle Consistency introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "Viewing Graph Solvability via Cycle Consistency is an award-winning computer vision paper.", "motivation": "Viewing Graph Solvability via Cycle Consistency targets a central problem recognized by the conference awards committee.", "implementation": "Viewing Graph Solvability via Cycle Consistency introduces a method or system selected as award-winning work.", "application": "Viewing Graph Solvability via Cycle Consistency is relevant to its conference track and downstream computer vision applications.", "problem": "Viewing Graph Solvability via Cycle Consistency targets a central problem recognized by the conference awards committee.", "contribution": "Viewing Graph Solvability via Cycle Consistency introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "ICCV 2021 Honorable Mention: Viewing Graph Solvability via Cycle Consistency is an award-winning computer vision paper.", "whyFollow": "ICCV 2021 Honorable Mention，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "Viewing Graph Solvability via Cycle Consistency is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "ICCV", "Honorable Mention", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "ICCV", "year": 2021, "award": "Honorable Mention", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "Common Objects in 3D: Large-Scale Learning and Evaluation of Real-life 3D Category Reconstruction", "authors": "Jeremy Reizenstein, Roman Shapovalov, Philipp Henzler, Luca Sbordone, Patrick Labatut, David Novotny", "venue": "ICCV", "year": 2021, "topic": "3D / Geometry / Reconstruction", "modality": "3D / Geometry / Reconstruction", "status": "CV 顶会 Best Paper 合集", "priority": "Medium", "task": "3D / Geometry / Reconstruction", "method": "Common Objects in 3D: Large-Scale Learning and Evaluation of Real-life 3D Category Reconstruction introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "Common Objects in 3D: Large-Scale Learning and Evaluation of Real-life 3D Category Reconstruction is an award-winning computer vision paper.", "motivation": "Common Objects in 3D: Large-Scale Learning and Evaluation of Real-life 3D Category Reconstruction targets a central problem recognized by the conference awards committee.", "implementation": "Common Objects in 3D: Large-Scale Learning and Evaluation of Real-life 3D Category Reconstruction introduces a method or system selected as award-winning work.", "application": "Common Objects in 3D: Large-Scale Learning and Evaluation of Real-life 3D Category Reconstruction is relevant to its conference track and downstream computer vision applications.", "problem": "Common Objects in 3D: Large-Scale Learning and Evaluation of Real-life 3D Category Reconstruction targets a central problem recognized by the conference awards committee.", "contribution": "Common Objects in 3D: Large-Scale Learning and Evaluation of Real-life 3D Category Reconstruction introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "ICCV 2021 Honorable Mention: Common Objects in 3D: Large-Scale Learning and Evaluation of Real-life 3D Category Reconstruction is an award-winning computer vision paper.", "whyFollow": "ICCV 2021 Honorable Mention，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "Common Objects in 3D: Large-Scale Learning and Evaluation of Real-life 3D Category Reconstruction is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "ICCV", "Honorable Mention", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "ICCV", "year": 2021, "award": "Honorable Mention", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "SinGAN: Learning a Generative Model from a Single Natural Image", "authors": "Tamar Rott Shaham, Tali Dekel, Tomer Michaeli", "venue": "ICCV", "year": 2019, "topic": "Generative Vision", "modality": "Generative Vision", "status": "CV 顶会 Best Paper 合集", "priority": "High", "task": "Generative Vision", "method": "SinGAN: Learning a Generative Model from a Single Natural Image introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "SinGAN: Learning a Generative Model from a Single Natural Image is an award-winning computer vision paper.", "motivation": "SinGAN: Learning a Generative Model from a Single Natural Image targets a central problem recognized by the conference awards committee.", "implementation": "SinGAN: Learning a Generative Model from a Single Natural Image introduces a method or system selected as award-winning work.", "application": "SinGAN: Learning a Generative Model from a Single Natural Image is relevant to its conference track and downstream computer vision applications.", "problem": "SinGAN: Learning a Generative Model from a Single Natural Image targets a central problem recognized by the conference awards committee.", "contribution": "SinGAN: Learning a Generative Model from a Single Natural Image introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "ICCV 2019 Best Paper: SinGAN: Learning a Generative Model from a Single Natural Image is an award-winning computer vision paper.", "whyFollow": "ICCV 2019 Best Paper，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "SinGAN: Learning a Generative Model from a Single Natural Image is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "ICCV", "Best Paper", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "ICCV", "year": 2019, "award": "Best Paper", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "PLMP: Point-Line Minimal Problems in Complete Multi-View Visibility", "authors": "Timothy Duff, Kathlen Kohn, Anton Leykin, Tomas Pajdla", "venue": "ICCV", "year": 2019, "topic": "3D / Geometry / Reconstruction", "modality": "3D / Geometry / Reconstruction", "status": "CV 顶会 Best Paper 合集", "priority": "High", "task": "3D / Geometry / Reconstruction", "method": "PLMP: Point-Line Minimal Problems in Complete Multi-View Visibility introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "PLMP: Point-Line Minimal Problems in Complete Multi-View Visibility is an award-winning computer vision paper.", "motivation": "PLMP: Point-Line Minimal Problems in Complete Multi-View Visibility targets a central problem recognized by the conference awards committee.", "implementation": "PLMP: Point-Line Minimal Problems in Complete Multi-View Visibility introduces a method or system selected as award-winning work.", "application": "PLMP: Point-Line Minimal Problems in Complete Multi-View Visibility is relevant to its conference track and downstream computer vision applications.", "problem": "PLMP: Point-Line Minimal Problems in Complete Multi-View Visibility targets a central problem recognized by the conference awards committee.", "contribution": "PLMP: Point-Line Minimal Problems in Complete Multi-View Visibility introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "ICCV 2019 Best Student Paper: PLMP: Point-Line Minimal Problems in Complete Multi-View Visibility is an award-winning computer vision paper.", "whyFollow": "ICCV 2019 Best Student Paper，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "PLMP: Point-Line Minimal Problems in Complete Multi-View Visibility is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "ICCV", "Best Student", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "ICCV", "year": 2019, "award": "Best Student Paper", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "Asynchronous Single-Photon 3D Imaging", "authors": "Anant Gupta, Atul Ingle, Mohit Gupta", "venue": "ICCV", "year": 2019, "topic": "3D / Geometry / Reconstruction", "modality": "3D / Geometry / Reconstruction", "status": "CV 顶会 Best Paper 合集", "priority": "Medium", "task": "3D / Geometry / Reconstruction", "method": "Asynchronous Single-Photon 3D Imaging introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "Asynchronous Single-Photon 3D Imaging is an award-winning computer vision paper.", "motivation": "Asynchronous Single-Photon 3D Imaging targets a central problem recognized by the conference awards committee.", "implementation": "Asynchronous Single-Photon 3D Imaging introduces a method or system selected as award-winning work.", "application": "Asynchronous Single-Photon 3D Imaging is relevant to its conference track and downstream computer vision applications.", "problem": "Asynchronous Single-Photon 3D Imaging targets a central problem recognized by the conference awards committee.", "contribution": "Asynchronous Single-Photon 3D Imaging introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "ICCV 2019 Honorable Mention: Asynchronous Single-Photon 3D Imaging is an award-winning computer vision paper.", "whyFollow": "ICCV 2019 Honorable Mention，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "Asynchronous Single-Photon 3D Imaging is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "ICCV", "Honorable Mention", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "ICCV", "year": 2019, "award": "Honorable Mention", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "Specifying Object Attributes and Relations in Interactive Scene Generation", "authors": "Oron Ashual, Lior Wolf", "venue": "ICCV", "year": 2019, "topic": "Generative Vision", "modality": "Generative Vision", "status": "CV 顶会 Best Paper 合集", "priority": "Medium", "task": "Generative Vision", "method": "Specifying Object Attributes and Relations in Interactive Scene Generation introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "Specifying Object Attributes and Relations in Interactive Scene Generation is an award-winning computer vision paper.", "motivation": "Specifying Object Attributes and Relations in Interactive Scene Generation targets a central problem recognized by the conference awards committee.", "implementation": "Specifying Object Attributes and Relations in Interactive Scene Generation introduces a method or system selected as award-winning work.", "application": "Specifying Object Attributes and Relations in Interactive Scene Generation is relevant to its conference track and downstream computer vision applications.", "problem": "Specifying Object Attributes and Relations in Interactive Scene Generation targets a central problem recognized by the conference awards committee.", "contribution": "Specifying Object Attributes and Relations in Interactive Scene Generation introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "ICCV 2019 Honorable Mention: Specifying Object Attributes and Relations in Interactive Scene Generation is an award-winning computer vision paper.", "whyFollow": "ICCV 2019 Honorable Mention，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "Specifying Object Attributes and Relations in Interactive Scene Generation is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "ICCV", "Honorable Mention", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "ICCV", "year": 2019, "award": "Honorable Mention", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "Mask R-CNN", "authors": "Kaiming He, Georgia Gkioxari, Piotr Dollar, Ross Girshick", "venue": "ICCV", "year": 2017, "topic": "Award-Winning Computer Vision", "modality": "Award-Winning Computer Vision", "status": "CV 顶会 Best Paper 合集", "priority": "High", "task": "Award-Winning Computer Vision", "method": "Mask R-CNN introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "Mask R-CNN is an award-winning computer vision paper.", "motivation": "Mask R-CNN targets a central problem recognized by the conference awards committee.", "implementation": "Mask R-CNN introduces a method or system selected as award-winning work.", "application": "Mask R-CNN is relevant to its conference track and downstream computer vision applications.", "problem": "Mask R-CNN targets a central problem recognized by the conference awards committee.", "contribution": "Mask R-CNN introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "ICCV 2017 Best Paper: Mask R-CNN is an award-winning computer vision paper.", "whyFollow": "ICCV 2017 Best Paper，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "Mask R-CNN is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "ICCV", "Best Paper", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "ICCV", "year": 2017, "award": "Best Paper", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "First Person Activity Forecasting with Online Inverse Reinforcement Learning", "authors": "Nicholas Rhinehart, Kris M. Kitani", "venue": "ICCV", "year": 2017, "topic": "Video / Embodied Perception", "modality": "Video / Embodied Perception", "status": "CV 顶会 Best Paper 合集", "priority": "Medium", "task": "Video / Embodied Perception", "method": "First Person Activity Forecasting with Online Inverse Reinforcement Learning introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "First Person Activity Forecasting with Online Inverse Reinforcement Learning is an award-winning computer vision paper.", "motivation": "First Person Activity Forecasting with Online Inverse Reinforcement Learning targets a central problem recognized by the conference awards committee.", "implementation": "First Person Activity Forecasting with Online Inverse Reinforcement Learning introduces a method or system selected as award-winning work.", "application": "First Person Activity Forecasting with Online Inverse Reinforcement Learning is relevant to its conference track and downstream computer vision applications.", "problem": "First Person Activity Forecasting with Online Inverse Reinforcement Learning targets a central problem recognized by the conference awards committee.", "contribution": "First Person Activity Forecasting with Online Inverse Reinforcement Learning introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "ICCV 2017 Honorable Mention: First Person Activity Forecasting with Online Inverse Reinforcement Learning is an award-winning computer vision paper.", "whyFollow": "ICCV 2017 Honorable Mention，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "First Person Activity Forecasting with Online Inverse Reinforcement Learning is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "ICCV", "Honorable Mention", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "ICCV", "year": 2017, "award": "Honorable Mention", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "Open Set Domain Adaptation", "authors": "Pau Panareda Busto, Juergen Gall", "venue": "ICCV", "year": 2017, "topic": "Award-Winning Computer Vision", "modality": "Award-Winning Computer Vision", "status": "CV 顶会 Best Paper 合集", "priority": "Medium", "task": "Award-Winning Computer Vision", "method": "Open Set Domain Adaptation introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "Open Set Domain Adaptation is an award-winning computer vision paper.", "motivation": "Open Set Domain Adaptation targets a central problem recognized by the conference awards committee.", "implementation": "Open Set Domain Adaptation introduces a method or system selected as award-winning work.", "application": "Open Set Domain Adaptation is relevant to its conference track and downstream computer vision applications.", "problem": "Open Set Domain Adaptation targets a central problem recognized by the conference awards committee.", "contribution": "Open Set Domain Adaptation introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "ICCV 2017 Honorable Mention: Open Set Domain Adaptation is an award-winning computer vision paper.", "whyFollow": "ICCV 2017 Honorable Mention，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "Open Set Domain Adaptation is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "ICCV", "Honorable Mention", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "ICCV", "year": 2017, "award": "Honorable Mention", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "Globally-Optimal Inlier Set Maximisation for Simultaneous Camera Pose and Feature Correspondence", "authors": "Dylan Campbell, Lars Petersson, Laurent Kneip, Hongdong Li", "venue": "ICCV", "year": 2017, "topic": "3D / Geometry / Reconstruction", "modality": "3D / Geometry / Reconstruction", "status": "CV 顶会 Best Paper 合集", "priority": "Medium", "task": "3D / Geometry / Reconstruction", "method": "Globally-Optimal Inlier Set Maximisation for Simultaneous Camera Pose and Feature Correspondence introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "Globally-Optimal Inlier Set Maximisation for Simultaneous Camera Pose and Feature Correspondence is an award-winning computer vision paper.", "motivation": "Globally-Optimal Inlier Set Maximisation for Simultaneous Camera Pose and Feature Correspondence targets a central problem recognized by the conference awards committee.", "implementation": "Globally-Optimal Inlier Set Maximisation for Simultaneous Camera Pose and Feature Correspondence introduces a method or system selected as award-winning work.", "application": "Globally-Optimal Inlier Set Maximisation for Simultaneous Camera Pose and Feature Correspondence is relevant to its conference track and downstream computer vision applications.", "problem": "Globally-Optimal Inlier Set Maximisation for Simultaneous Camera Pose and Feature Correspondence targets a central problem recognized by the conference awards committee.", "contribution": "Globally-Optimal Inlier Set Maximisation for Simultaneous Camera Pose and Feature Correspondence introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "ICCV 2017 Honorable Mention: Globally-Optimal Inlier Set Maximisation for Simultaneous Camera Pose and Feature Correspondence is an award-winning computer vision paper.", "whyFollow": "ICCV 2017 Honorable Mention，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "Globally-Optimal Inlier Set Maximisation for Simultaneous Camera Pose and Feature Correspondence is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "ICCV", "Honorable Mention", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "ICCV", "year": 2017, "award": "Honorable Mention", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "Minimalist Vision with Freeform Pixels", "authors": "Jeremy Klotz, Shree Nayar", "venue": "ECCV", "year": 2024, "topic": "Award-Winning Computer Vision", "modality": "Award-Winning Computer Vision", "status": "CV 顶会 Best Paper 合集", "priority": "High", "task": "Award-Winning Computer Vision", "method": "Minimalist Vision with Freeform Pixels introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "Minimalist Vision with Freeform Pixels is an award-winning computer vision paper.", "motivation": "Minimalist Vision with Freeform Pixels targets a central problem recognized by the conference awards committee.", "implementation": "Minimalist Vision with Freeform Pixels introduces a method or system selected as award-winning work.", "application": "Minimalist Vision with Freeform Pixels is relevant to its conference track and downstream computer vision applications.", "problem": "Minimalist Vision with Freeform Pixels targets a central problem recognized by the conference awards committee.", "contribution": "Minimalist Vision with Freeform Pixels introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "ECCV 2024 Best Paper: Minimalist Vision with Freeform Pixels is an award-winning computer vision paper.", "whyFollow": "ECCV 2024 Best Paper，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "Minimalist Vision with Freeform Pixels is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "ECCV", "Best Paper", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "ECCV", "year": 2024, "award": "Best Paper", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "Rasterized Edge Gradients: Handling Discontinuities Differentially", "authors": "Stanislav Pidhorskyi, Tomas Simon, Gabriel Schwartz, He Wen, Yaser Sheikh, Jason Saragih", "venue": "ECCV", "year": 2024, "topic": "Segmentation", "modality": "Segmentation", "status": "CV 顶会 Best Paper 合集", "priority": "Medium", "task": "Segmentation", "method": "Rasterized Edge Gradients: Handling Discontinuities Differentially introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "Rasterized Edge Gradients: Handling Discontinuities Differentially is an award-winning computer vision paper.", "motivation": "Rasterized Edge Gradients: Handling Discontinuities Differentially targets a central problem recognized by the conference awards committee.", "implementation": "Rasterized Edge Gradients: Handling Discontinuities Differentially introduces a method or system selected as award-winning work.", "application": "Rasterized Edge Gradients: Handling Discontinuities Differentially is relevant to its conference track and downstream computer vision applications.", "problem": "Rasterized Edge Gradients: Handling Discontinuities Differentially targets a central problem recognized by the conference awards committee.", "contribution": "Rasterized Edge Gradients: Handling Discontinuities Differentially introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "ECCV 2024 Honorable Mention: Rasterized Edge Gradients: Handling Discontinuities Differentially is an award-winning computer vision paper.", "whyFollow": "ECCV 2024 Honorable Mention，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "Rasterized Edge Gradients: Handling Discontinuities Differentially is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "ECCV", "Honorable Mention", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "ECCV", "year": 2024, "award": "Honorable Mention", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "Concept Arithmetics for Circumventing Concept Inhibition in Diffusion Models", "authors": "Vitali Petsiuk, Kate Saenko", "venue": "ECCV", "year": 2024, "topic": "Generative Vision", "modality": "Generative Vision", "status": "CV 顶会 Best Paper 合集", "priority": "Medium", "task": "Generative Vision", "method": "Concept Arithmetics for Circumventing Concept Inhibition in Diffusion Models introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "Concept Arithmetics for Circumventing Concept Inhibition in Diffusion Models is an award-winning computer vision paper.", "motivation": "Concept Arithmetics for Circumventing Concept Inhibition in Diffusion Models targets a central problem recognized by the conference awards committee.", "implementation": "Concept Arithmetics for Circumventing Concept Inhibition in Diffusion Models introduces a method or system selected as award-winning work.", "application": "Concept Arithmetics for Circumventing Concept Inhibition in Diffusion Models is relevant to its conference track and downstream computer vision applications.", "problem": "Concept Arithmetics for Circumventing Concept Inhibition in Diffusion Models targets a central problem recognized by the conference awards committee.", "contribution": "Concept Arithmetics for Circumventing Concept Inhibition in Diffusion Models introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "ECCV 2024 Honorable Mention: Concept Arithmetics for Circumventing Concept Inhibition in Diffusion Models is an award-winning computer vision paper.", "whyFollow": "ECCV 2024 Honorable Mention，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "Concept Arithmetics for Circumventing Concept Inhibition in Diffusion Models is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "ECCV", "Honorable Mention", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "ECCV", "year": 2024, "award": "Honorable Mention", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "On the Versatile Uses of Partial Distance Correlation in Deep Learning", "authors": "Xinghao Zhen, Zhun Deng, Rudrasis Chakraborty, Vikas Singh", "venue": "ECCV", "year": 2022, "topic": "Award-Winning Computer Vision", "modality": "Award-Winning Computer Vision", "status": "CV 顶会 Best Paper 合集", "priority": "High", "task": "Award-Winning Computer Vision", "method": "On the Versatile Uses of Partial Distance Correlation in Deep Learning introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "On the Versatile Uses of Partial Distance Correlation in Deep Learning is an award-winning computer vision paper.", "motivation": "On the Versatile Uses of Partial Distance Correlation in Deep Learning targets a central problem recognized by the conference awards committee.", "implementation": "On the Versatile Uses of Partial Distance Correlation in Deep Learning introduces a method or system selected as award-winning work.", "application": "On the Versatile Uses of Partial Distance Correlation in Deep Learning is relevant to its conference track and downstream computer vision applications.", "problem": "On the Versatile Uses of Partial Distance Correlation in Deep Learning targets a central problem recognized by the conference awards committee.", "contribution": "On the Versatile Uses of Partial Distance Correlation in Deep Learning introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "ECCV 2022 Best Paper: On the Versatile Uses of Partial Distance Correlation in Deep Learning is an award-winning computer vision paper.", "whyFollow": "ECCV 2022 Best Paper，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "On the Versatile Uses of Partial Distance Correlation in Deep Learning is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "ECCV", "Best Paper", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "ECCV", "year": 2022, "award": "Best Paper", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "Pose-NDF: Modeling Human Pose Manifolds with Neural Distance Fields", "authors": "Garvita Tiwari, Nikolaos Sarafianos, Tony Tung, Gerard Pons-Moll", "venue": "ECCV", "year": 2022, "topic": "3D / Geometry / Reconstruction", "modality": "3D / Geometry / Reconstruction", "status": "CV 顶会 Best Paper 合集", "priority": "Medium", "task": "3D / Geometry / Reconstruction", "method": "Pose-NDF: Modeling Human Pose Manifolds with Neural Distance Fields introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "Pose-NDF: Modeling Human Pose Manifolds with Neural Distance Fields is an award-winning computer vision paper.", "motivation": "Pose-NDF: Modeling Human Pose Manifolds with Neural Distance Fields targets a central problem recognized by the conference awards committee.", "implementation": "Pose-NDF: Modeling Human Pose Manifolds with Neural Distance Fields introduces a method or system selected as award-winning work.", "application": "Pose-NDF: Modeling Human Pose Manifolds with Neural Distance Fields is relevant to its conference track and downstream computer vision applications.", "problem": "Pose-NDF: Modeling Human Pose Manifolds with Neural Distance Fields targets a central problem recognized by the conference awards committee.", "contribution": "Pose-NDF: Modeling Human Pose Manifolds with Neural Distance Fields introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "ECCV 2022 Honorable Mention: Pose-NDF: Modeling Human Pose Manifolds with Neural Distance Fields is an award-winning computer vision paper.", "whyFollow": "ECCV 2022 Honorable Mention，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "Pose-NDF: Modeling Human Pose Manifolds with Neural Distance Fields is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "ECCV", "Honorable Mention", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "ECCV", "year": 2022, "award": "Honorable Mention", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "A Level Set Theory for Neural Implicit Evolution under Explicit Flows", "authors": "Ishit Mehta, Manmohan Chandraker, Ravi Ramamoorthi", "venue": "ECCV", "year": 2022, "topic": "Award-Winning Computer Vision", "modality": "Award-Winning Computer Vision", "status": "CV 顶会 Best Paper 合集", "priority": "Medium", "task": "Award-Winning Computer Vision", "method": "A Level Set Theory for Neural Implicit Evolution under Explicit Flows introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "A Level Set Theory for Neural Implicit Evolution under Explicit Flows is an award-winning computer vision paper.", "motivation": "A Level Set Theory for Neural Implicit Evolution under Explicit Flows targets a central problem recognized by the conference awards committee.", "implementation": "A Level Set Theory for Neural Implicit Evolution under Explicit Flows introduces a method or system selected as award-winning work.", "application": "A Level Set Theory for Neural Implicit Evolution under Explicit Flows is relevant to its conference track and downstream computer vision applications.", "problem": "A Level Set Theory for Neural Implicit Evolution under Explicit Flows targets a central problem recognized by the conference awards committee.", "contribution": "A Level Set Theory for Neural Implicit Evolution under Explicit Flows introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "ECCV 2022 Honorable Mention: A Level Set Theory for Neural Implicit Evolution under Explicit Flows is an award-winning computer vision paper.", "whyFollow": "ECCV 2022 Honorable Mention，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "A Level Set Theory for Neural Implicit Evolution under Explicit Flows is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "ECCV", "Honorable Mention", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "ECCV", "year": 2022, "award": "Honorable Mention", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "Towards Streaming Perception", "authors": "Mengtian Li, Yu-Xiong Wang, Deva Ramanan", "venue": "ECCV", "year": 2020, "topic": "Video / Embodied Perception", "modality": "Video / Embodied Perception", "status": "CV 顶会 Best Paper 合集", "priority": "Medium", "task": "Video / Embodied Perception", "method": "Towards Streaming Perception introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "Towards Streaming Perception is an award-winning computer vision paper.", "motivation": "Towards Streaming Perception targets a central problem recognized by the conference awards committee.", "implementation": "Towards Streaming Perception introduces a method or system selected as award-winning work.", "application": "Towards Streaming Perception is relevant to its conference track and downstream computer vision applications.", "problem": "Towards Streaming Perception targets a central problem recognized by the conference awards committee.", "contribution": "Towards Streaming Perception introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "ECCV 2020 Honorable Mention: Towards Streaming Perception is an award-winning computer vision paper.", "whyFollow": "ECCV 2020 Honorable Mention，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "Towards Streaming Perception is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "ECCV", "Honorable Mention", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "ECCV", "year": 2020, "award": "Honorable Mention", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis", "authors": "Ben Mildenhall, Pratul P. Srinivasan, Matthew Tancik, Jonathan T. Barron, Ravi Ramamoorthi, Ren Ng", "venue": "ECCV", "year": 2020, "topic": "3D / Geometry / Reconstruction", "modality": "3D / Geometry / Reconstruction", "status": "CV 顶会 Best Paper 合集", "priority": "Medium", "task": "3D / Geometry / Reconstruction", "method": "NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis is an award-winning computer vision paper.", "motivation": "NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis targets a central problem recognized by the conference awards committee.", "implementation": "NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis introduces a method or system selected as award-winning work.", "application": "NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis is relevant to its conference track and downstream computer vision applications.", "problem": "NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis targets a central problem recognized by the conference awards committee.", "contribution": "NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "ECCV 2020 Honorable Mention: NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis is an award-winning computer vision paper.", "whyFollow": "ECCV 2020 Honorable Mention，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "ECCV", "Honorable Mention", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "ECCV", "year": 2020, "award": "Honorable Mention", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "Implicit 3D Orientation Learning for 6D Object Detection from RGB Images", "authors": "Martin Sundermeyer, Zoltan-Csaba Marton, Maximilian Durner, Manuel Brucker, Rudolph Triebel", "venue": "ECCV", "year": 2018, "topic": "3D / Geometry / Reconstruction", "modality": "3D / Geometry / Reconstruction", "status": "CV 顶会 Best Paper 合集", "priority": "High", "task": "3D / Geometry / Reconstruction", "method": "Implicit 3D Orientation Learning for 6D Object Detection from RGB Images introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "Implicit 3D Orientation Learning for 6D Object Detection from RGB Images is an award-winning computer vision paper.", "motivation": "Implicit 3D Orientation Learning for 6D Object Detection from RGB Images targets a central problem recognized by the conference awards committee.", "implementation": "Implicit 3D Orientation Learning for 6D Object Detection from RGB Images introduces a method or system selected as award-winning work.", "application": "Implicit 3D Orientation Learning for 6D Object Detection from RGB Images is relevant to its conference track and downstream computer vision applications.", "problem": "Implicit 3D Orientation Learning for 6D Object Detection from RGB Images targets a central problem recognized by the conference awards committee.", "contribution": "Implicit 3D Orientation Learning for 6D Object Detection from RGB Images introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "ECCV 2018 Best Paper: Implicit 3D Orientation Learning for 6D Object Detection from RGB Images is an award-winning computer vision paper.", "whyFollow": "ECCV 2018 Best Paper，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "Implicit 3D Orientation Learning for 6D Object Detection from RGB Images is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "ECCV", "Best Paper", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "ECCV", "year": 2018, "award": "Best Paper", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "Group Normalization", "authors": "Yuxin Wu, Kaiming He", "venue": "ECCV", "year": 2018, "topic": "Vision Backbones", "modality": "Vision Backbones", "status": "CV 顶会 Best Paper 合集", "priority": "Medium", "task": "Vision Backbones", "method": "Group Normalization introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "Group Normalization is an award-winning computer vision paper.", "motivation": "Group Normalization targets a central problem recognized by the conference awards committee.", "implementation": "Group Normalization introduces a method or system selected as award-winning work.", "application": "Group Normalization is relevant to its conference track and downstream computer vision applications.", "problem": "Group Normalization targets a central problem recognized by the conference awards committee.", "contribution": "Group Normalization introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "ECCV 2018 Honorable Mention: Group Normalization is an award-winning computer vision paper.", "whyFollow": "ECCV 2018 Honorable Mention，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "Group Normalization is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "ECCV", "Honorable Mention", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "ECCV", "year": 2018, "award": "Honorable Mention", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "GANimation: Anatomically-aware Facial Animation from a Single Image", "authors": "Albert Pumarola, Antonio Agudo, Aleix M. Martinez, Alberto Sanfeliu, Francesc Moreno-Noguer", "venue": "ECCV", "year": 2018, "topic": "Generative Vision", "modality": "Generative Vision", "status": "CV 顶会 Best Paper 合集", "priority": "Medium", "task": "Generative Vision", "method": "GANimation: Anatomically-aware Facial Animation from a Single Image introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "GANimation: Anatomically-aware Facial Animation from a Single Image is an award-winning computer vision paper.", "motivation": "GANimation: Anatomically-aware Facial Animation from a Single Image targets a central problem recognized by the conference awards committee.", "implementation": "GANimation: Anatomically-aware Facial Animation from a Single Image introduces a method or system selected as award-winning work.", "application": "GANimation: Anatomically-aware Facial Animation from a Single Image is relevant to its conference track and downstream computer vision applications.", "problem": "GANimation: Anatomically-aware Facial Animation from a Single Image targets a central problem recognized by the conference awards committee.", "contribution": "GANimation: Anatomically-aware Facial Animation from a Single Image introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "ECCV 2018 Honorable Mention: GANimation: Anatomically-aware Facial Animation from a Single Image is an award-winning computer vision paper.", "whyFollow": "ECCV 2018 Honorable Mention，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "GANimation: Anatomically-aware Facial Animation from a Single Image is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "ECCV", "Honorable Mention", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "ECCV", "year": 2018, "award": "Honorable Mention", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "Real-Time 3D Reconstruction and 6-DoF Tracking with an Event Camera", "authors": "Hanme Kim, Stefan Leutenegger, Andrew J. Davison", "venue": "ECCV", "year": 2016, "topic": "3D / Geometry / Reconstruction", "modality": "3D / Geometry / Reconstruction", "status": "CV 顶会 Best Paper 合集", "priority": "High", "task": "3D / Geometry / Reconstruction", "method": "Real-Time 3D Reconstruction and 6-DoF Tracking with an Event Camera introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "Real-Time 3D Reconstruction and 6-DoF Tracking with an Event Camera is an award-winning computer vision paper.", "motivation": "Real-Time 3D Reconstruction and 6-DoF Tracking with an Event Camera targets a central problem recognized by the conference awards committee.", "implementation": "Real-Time 3D Reconstruction and 6-DoF Tracking with an Event Camera introduces a method or system selected as award-winning work.", "application": "Real-Time 3D Reconstruction and 6-DoF Tracking with an Event Camera is relevant to its conference track and downstream computer vision applications.", "problem": "Real-Time 3D Reconstruction and 6-DoF Tracking with an Event Camera targets a central problem recognized by the conference awards committee.", "contribution": "Real-Time 3D Reconstruction and 6-DoF Tracking with an Event Camera introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "ECCV 2016 Best Paper: Real-Time 3D Reconstruction and 6-DoF Tracking with an Event Camera is an award-winning computer vision paper.", "whyFollow": "ECCV 2016 Best Paper，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "Real-Time 3D Reconstruction and 6-DoF Tracking with an Event Camera is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "ECCV", "Best Paper", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "ECCV", "year": 2016, "award": "Best Paper", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" }, { "title": "The Fast Bilateral Solver", "authors": "Jonathan T. Barron, Ben Poole", "venue": "ECCV", "year": 2016, "topic": "Award-Winning Computer Vision", "modality": "Award-Winning Computer Vision", "status": "CV 顶会 Best Paper 合集", "priority": "Medium", "task": "Award-Winning Computer Vision", "method": "The Fast Bilateral Solver introduces a method or system selected as award-winning work.", "hardware": "公开论文 / 官方获奖列表", "dataset": "见论文实验设置", "abstract": "The Fast Bilateral Solver is an award-winning computer vision paper.", "motivation": "The Fast Bilateral Solver targets a central problem recognized by the conference awards committee.", "implementation": "The Fast Bilateral Solver introduces a method or system selected as award-winning work.", "application": "The Fast Bilateral Solver is relevant to its conference track and downstream computer vision applications.", "problem": "The Fast Bilateral Solver targets a central problem recognized by the conference awards committee.", "contribution": "The Fast Bilateral Solver introduces a method or system selected as award-winning work.", "limitation": "获奖代表质量和影响力，但具体适用边界仍需结合任务、数据和后续工作判断。", "insight": "ECCV 2016 Honorable Mention: The Fast Bilateral Solver is an award-winning computer vision paper.", "whyFollow": "ECCV 2016 Honorable Mention，适合用于建立 CV 顶会高质量论文阅读路线。", "openQuestion": "后续可继续补充复现状态、开源代码质量和与计算成像方向的连接。", "directionNote": "The Fast Bilateral Solver is relevant to its conference track and downstream computer vision applications.", "tags": [ "award-winning", "ECCV", "Honorable Mention", "CVPR/ICCV/ECCV" ], "code": { "available": false, "url": "" }, "url": "", "updated": "2026-05-22", "awards": [ { "conference": "ECCV", "year": 2016, "award": "Honorable Mention", "source": "CVF Computer Vision Awards", "sourceUrl": "https://www.thecvf.com/?page_id=413" } ], "citationCount": -1, "citationSource": "", "semanticScholarUrl": "", "doi": "", "arxiv": "" } ]