️ 顶级学术会议会议全称级别每年论文数官网开放获取CVPRIEEE/CVF Computer Vision and Pattern RecognitionCCF-A, 顶会~2400https://cvpr.thecvf.comhttps://openaccess.thecvf.com/CVPR2025ICCVIEEE/CVF International Conference on Computer VisionCCF-A, 顶会~1800https://iccv2025.thecvf.comhttps://openaccess.thecvf.com/ICCV2025ECCVEuropean Conference on Computer VisionCCF-B~1600https://eccv2026.org部分开放NeurIPSNeural Information Processing SystemsCCF-A~3200https://neurips.cchttps://papers.nips.ccICLRInternational Conference on Learning RepresentationsCCF-A~2300https://iclr.cc开放ACM MMACM MultimediaCCF-A~1300https://www.acmmm.org部分AAAIAAAI Conference on Artificial IntelligenceCCF-A~2300https://aaai.org部分BMVCBritish Machine Vision ConferenceCCF-C~600https://bmvc2025.org开放WACVWinter Conference on Applications of CV-~800https://wacv2025.thecvf.com开放 顶级期刊期刊影响因子 (2024)PublisherTPAMI(IEEE Trans. on Pattern Analysis and Machine Intelligence)~24.5IEEEIJCV(International Journal of Computer Vision)~19.5SpringerTIP(IEEE Trans. on Image Processing)~10.8IEEECVIU(Computer Vision and Image Understanding)~7.6ElsevierPR(Pattern Recognition)~8.5ElsevierMIA(Medical Image Analysis)~13.8Elsevier 顶级实验室实验室所属机构知名教授方向SAIL(Stanford AI Lab)StanfordFei-Fei Li, Leonidas GuibasCV MLCSAILMITBill Freeman, Antonio TorralbaCV GraphicsBAIRUC BerkeleyAlyosha Efros, Trevor DarrellCV NLPVGG / OIIOxfordAndrew Zisserman, Andrea Vedaldi识别/视频INRIA (Willow)INRIA / ENSJean Ponce, Josef Sivic3D视觉MPI InformaticsMax PlanckBernt Schiele, Christian Theobalt人体/3DFacebook AI Research (FAIR)MetaYann LeCun, Michael Tschannen自监督/基础模型Google DeepMindGoogleNando de Freitas多模态/RLCVLETH ZurichLuc Van Gool3D/自动驾驶CMU RICMUTakeo Kanade, Martial Hebert机器人视觉商汤研究院SenseTime汤晓鸥全栈CV旷视研究院Megvii孙剑检测识别腾讯优图Tencent贾佳亚应用CV阿里巴巴达摩院Alibaba团队视觉AI百度视觉Baidu团队自动驾驶/OCR GitHub Awesome 系列仓库链接内容Awesome Computer Visionhttps://github.com/jbhuang0604/awesome-computer-vision综合CV资源列表Awesome Deep Visionhttps://github.com/kjw0612/awesome-deep-vision深度学习CVAwesome Object Detectionhttps://github.com/amusi/awesome-object-detection目标检测论文/代码Awesome 3D Reconstructionhttps://github.com/openMVG/awesome_3DReconstruction3D重建Awesome SLAMSLAM相关同步定位与建图Awesome NeRFhttps://github.com/awesome-NeRF/awesome-NeRFNeRF资源大全Awesome Vision TransformerVision Transformer相关ViT论文/代码Awesome Face Recognition人脸识别相关人脸识别文献Awesome Self-Supervised Learninghttps://github.com/jason718/awesome-self-supervised-learning自监督学习Awesome Diffusion Modelshttps://github.com/hee9joon/Awesome-Diffusion-Models扩散模型Papers With Codehttps://paperswithcode.com/论文代码BenchmarkSOTA Modelshttps://paperswithcode.com/sota各任务SOTA排行榜 关键数据集图像分类数据集图片数类别数用途链接ImageNet-1K1.2M1000通用分类ImageNetCIFAR-10/10060K10/100小规模分类https://www.cs.toronto.edu/~kriz/cifar.htmlMNIST70K10手写数字Index of /exdb/mnistFashion-MNIST70K10时尚分类GitHub - zalandoresearch/fashion-mnist: A MNIST-like fashion product database. Benchmark · GitHubPlaces3651.8M365场景分类http://places2.csail.mit.edu/目标检测数据集图片数类别数标注方式链接COCO330K80bboxsegCOCO - Common Objects in ContextPASCAL VOC11.5K20bboxsegThe PASCAL Visual Object Classes HomepageOpen Images9M600bboxseghttps://storage.googleapis.com/openimages/web/index.htmlVisDrone86K10无人机视角https://github.com/VisDrone/VisDrone-DatasetDOTA280K15旋转框https://captain-whu.github.io/DOTA/图像分割数据集任务链接ADE20K语义分割https://groups.csail.mit.edu/vision/datasets/ADE20K/Cityscapes街景分割https://www.cityscapes-dataset.com/Pascal Context场景理解http://www.cs.stanford.edu/~roozbeh/pascal-context/KITTI自动驾驶The KITTI Vision Benchmark SuiteCOCO-Stuff语义分割GitHub - nightrome/cocostuff: The official homepage of the COCO-Stuff dataset. · GitHub人脸数据集数量链接LFW13Khttp://vis-www.cs.umass.edu/lfw/MS-Celeb-1M10MMS-Celeb-1M: Challenge of Recognizing One Million Celebrities in the Real World - Microsoft ResearchVGGFace23.3MVisual Geometry Group - University of OxfordMegaFace1MMegaFace3D视觉数据集用途链接ShapeNet3D形状ShapeNetModelNet3D分类Princeton ModelNetScanNet室内3DScanNet | Richly-annotated 3D Reconstructions of Indoor ScenesKITTI自动驾驶3DThe KITTI Vision Benchmark SuiteWaymo Open自动驾驶https://waymo.com/open/nuScenes自动驾驶https://www.nuscenes.org/MegaDepth3D重建https://www.cs.cornell.edu/projects/megadepth/ 开发工具与框架核心框架PyTorch: https://pytorch.org/TensorFlow: https://www.tensorflow.org/JAX: GitHub - jax-ml/jax: Composable transformations of PythonNumPy programs: differentiate, vectorize, JIT to GPU/TPU, and more · GitHubKeras: Keras: Deep Learning for humansCV专用框架框架链接特点OpenCVOpenCV - Open Computer Vision Library经典CV库MMDetectionGitHub - open-mmlab/mmdetection: OpenMMLab Detection Toolbox and Benchmark · GitHub检测统一框架MMSegmentationGitHub - open-mmlab/mmsegmentation: OpenMMLab Semantic Segmentation Toolbox and Benchmark. · GitHub分割统一框架MMPoseGitHub - open-mmlab/mmpose: OpenMMLab Pose Estimation Toolbox and Benchmark. · GitHub姿态估计MMOCRhttps://github.com/open-mmlab/mmocrOCR工具箱MMTrackinghttps://github.com/open-mmlab/mmtracking目标跟踪Detectron2https://github.com/facebookresearch/detectron2Meta检测框架Segmentation Modelshttps://github.com/qubvel-org/segmentation_models.pytorchPyTorch分割TIMMhttps://github.com/huggingface/pytorch-image-models预训练模型库PyTorch3DGitHub - facebookresearch/pytorch3d: PyTorch3D is FAIRs library of reusable components for deep learning with 3D data · GitHub3D深度学习KorniaGitHub - kornia/kornia: Geometric Computer Vision Library for Spatial AI · GitHub可微分CV库FiftyOneGitHub - voxel51/fiftyone: Refine high-quality datasets and visual AI models · GitHub数据集可视化工具部署工具工具链接用途ONNXONNX | Home模型中间表示TensorRTTensorRT SDK | NVIDIA DeveloperNVIDIA推理优化OpenVINOhttps://docs.openvino.ai/Intel推理引擎TFLitehttps://www.tensorflow.org/lite移动端部署NCNNGitHub - Tencent/ncnn: ncnn is a high-performance neural network inference framework optimized for the mobile platform · GitHub腾讯移动端推理MNNGitHub - alibaba/MNN: MNN: A blazing-fast, lightweight inference engine battle-tested by Alibaba, powering high-performance on-device LLMs and Edge AI. · GitHub阿里巴巴推理框架TVMApache TVM端到端编译栈 学习平台与社区平台链接特点Papers With Codehttps://paperswithcode.com/论文代码排行榜Hugging Facehttps://huggingface.co/models模型库数据集SpacesKagglehttps://www.kaggle.com/CV竞赛数据集NotebookGitHubhttps://github.com/源代码复现Google Colabhttps://colab.research.google.com/免费GPURoboflowhttps://roboflow.com/CV数据管理与微调CVF开放访问https://openaccess.thecvf.com/CVPR/ICCV全文arXivhttps://arxiv.org/list/cs.CV/recent最新CV论文Google Scholarhttps://scholar.google.com/引文搜索知乎https://www.zhihu.com/topic/19551392中文CV讨论 论文下载与搜索工具链接arXiv (cs.CV)https://arxiv.org/list/cs.CV/recentarXiv Sanityhttps://arxiv-sanity-lite.com/ (Andrej Karpathy)Papers With Codehttps://paperswithcode.com/CVF Open Accesshttps://openaccess.thecvf.com/Semantic ScholarSemantic Scholar | AI-Powered Research ToolGoogle Scholarhttps://scholar.google.com/Connected Papershttps://www.connectedpapers.com/ResearchRabbitResearchRabbit: AI Tool for Smarter, Faster Literature ReviewsSci-Hub非必要不使用Aminerhttps://www.aminer.cn/ 2024-2026 前沿趋势方向代表性工作趋势说明视觉基础模型SAM 2, DINOv2, ViT-22B大一统模型few-shot/zero-shot多模态大模型GPT-4V, Gemini, LLaVA, Qwen-VL视觉语言融合3D生成3D Gaussian Splatting, Zero-1-to-33D AIGC爆发视频生成Sora, VideoPoet, Stable Video文生视频/图生视频可微渲染NeRF, 3DGS, Flash3D隐式/显式3D表示自监督DINOv2, I-JEPA, MAE减少对标注的依赖扩散模型SD3, Flux, DDPM图像生成主流范式世界模型UniSim, DayDreamer视觉物理仿真平铺CNN/ConvNeXtConvNeXt V2CNN复兴媲美ViT端侧部署MobileNetV4, EfficientViT手机上跑大模型对抗鲁棒性PGD-AT, TRADES, Random Smoothing让CV模型更安全可解释性Grad-CAM, LIME, SHAP, Attention Rollout理解模型决策原因 最后更新: 2026-06-17新增对抗攻击与防御、可解释性模块附录深层补充一、2024-2026 最新顶会论文获取渠道1.1 顶会论文搜索技巧渠道网址使用技巧CVF Open Accesshttps://openaccess.thecvf.com统一搜索 CVPR/ICCV/ECCV/WACV 全部论文全文Dblphttps://dblp.org按会议年份过滤如dblp.org/db/conf/cvpr/cvpr2025.htmlarXiv cs.CVComputer Vision and Pattern RecognitionRSS 订阅 按更新时间追踪最新论文Semantic Scholarhttps://www.semanticscholar.orgAPI 批量搜索结构化参考文献Google Scholarhttps://scholar.google.com引文追踪作者主页设置提醒新论文自动邮件通知1.2 Papers with Code 的高级筛选Papers with Code 不仅仅是论文代码的配对还提供强大的筛选功能按任务筛选paperswithcode.com/task/object-detection按不同 benchmark 查看 SOTA每个 benchmark 有 leaderboard 展示 Top 10 方法和数值可查看历史 SOTA 变迁按时间线的精度提升曲线按数据集筛选paperswithcode.com/datasets查看数据集页面每个数据集展示其相关的 benchmark 和 SOTA 方法按方法筛选支持搜索特定方法如 DETR和对比不同方法的 benchmark 表现RSS / Watchlist订阅特定任务/数据集新 SOTA 出现时自动通知1.3 Hugging Face Daily Papers每日前 5-10 篇最受关注的 arXiv 论文由社区共同投票upvote决定每天更新覆盖 CV / NLP / 多模态 / 语音 / 强化学习配合 Hugging Face 的 model card 可直接试验链接https://huggingface.co/papers二、实用的研究工具2.1 Connected Papers论文关系图Connected Papers | Find and explore academic papers输入一篇论文 → 生成论文影响力图谱节点论文越大代表引用越多或相关性越高边引用关系共被引、相关引用的强引用关系颜色按时间越深色越新功能Prior Works该论文基于的早期工作 Derivative Works后续发展用途快速了解一个领域的边界和关键路径发现意料之外的跨领域联系为综述提供系统化的文献框架2.2 Semantic Scholar研究搜索https://www.semanticscholar.org/AI 驱动的学术搜索引擎TLDRToo Long; Didnt ReadAI 生成的一句话论文摘要Citation Graph引用网络的结构化展示Influential Citations标记出最有影响力的引用API支持批量论文搜索和元数据获取Research Feed基于你的阅读历史推荐新论文2.3 Arxiv Sanity论文推荐https://arxiv-sanity-lite.com/Andrej Karpathy 维护从 arXiv 最新论文中根据你的偏好推荐通过点赞和收藏学习你的兴趣每周邮件摘要比直接刷 arXiv 高效得多2.4 科研 AI 助手工具功能链接Consensus检索一句话的科学答案附带引用https://consensus.appGalactica开源面向科研的 LLM支持公式/代码https://galactica.orgElicit自动化文献综述搜索和对比https://elicit.comSciSpace论文解读助手解释公式/图表https://typeset.ioPerplexity Pro实时搜索合成科研答案https://perplexity.ai三、中文社区资源3.1 知乎 CV 话题知乎专栏/话题描述链接计算机视觉话题30 万关注核心讨论区zhihu.com/topic/19551392AI 漫游指北李沐的知乎专栏高质量论文解读zhihu.com/column/c_1251902395475275776CV 前沿论文速递每日/每周最新论文zhihu.com/search?qCVPR2025论文计算机视觉与深度学习黄浴老师的技术分析知乎搜索黄浴3.2 公众号推荐公众号特点机器之心每日 AI 新闻覆盖全面量子位产业技术趋势分析CVerCV 入门到竞赛中文教程最全OpenMMLabOpenMMLab 官方算法开源详解AI 科技评论业界深度访谈Datawhale公益学习组织组团学 CV极市平台算法竞赛、工业应用深度学习初学者NewBeeNLP侧重 CV/DL 学习路径3.3 国内技术博客博客地址OpenMMLab 技术博客https://github.com/open-mmlab/OpenMMLabBlogDatawhale 教程Datawhale · GitHub动手学深度学习李沐https://d2l.ai/中文版CS231n 中文笔记ShowMeAI知识社区GiantPandaCV 公众号知乎/GitHub 搜 GiantPandaCVB 站 CV 论文精讲搜索论文精讲 计算机视觉3.4 国内竞赛平台平台描述链接天池阿里CV 比赛 数据集 GPU 算力https://tianchi.aliyun.comKaggle CN百度Kaggle 中文版https://kaggle.com中文社区Biendata国内数据竞赛https://www.biendata.xyzDataFountain数据竞赛https://www.datafountain.cn
计算机视觉资源总索引
发布时间：2026/6/18 14:27:44
️ 顶级学术会议会议全称级别每年论文数官网开放获取CVPRIEEE/CVF Computer Vision and Pattern RecognitionCCF-A, 顶会~2400https://cvpr.thecvf.comhttps://openaccess.thecvf.com/CVPR2025ICCVIEEE/CVF International Conference on Computer VisionCCF-A, 顶会~1800https://iccv2025.thecvf.comhttps://openaccess.thecvf.com/ICCV2025ECCVEuropean Conference on Computer VisionCCF-B~1600https://eccv2026.org部分开放NeurIPSNeural Information Processing SystemsCCF-A~3200https://neurips.cchttps://papers.nips.ccICLRInternational Conference on Learning RepresentationsCCF-A~2300https://iclr.cc开放ACM MMACM MultimediaCCF-A~1300https://www.acmmm.org部分AAAIAAAI Conference on Artificial IntelligenceCCF-A~2300https://aaai.org部分BMVCBritish Machine Vision ConferenceCCF-C~600https://bmvc2025.org开放WACVWinter Conference on Applications of CV-~800https://wacv2025.thecvf.com开放 顶级期刊期刊影响因子 (2024)PublisherTPAMI(IEEE Trans. on Pattern Analysis and Machine Intelligence)~24.5IEEEIJCV(International Journal of Computer Vision)~19.5SpringerTIP(IEEE Trans. on Image Processing)~10.8IEEECVIU(Computer Vision and Image Understanding)~7.6ElsevierPR(Pattern Recognition)~8.5ElsevierMIA(Medical Image Analysis)~13.8Elsevier 顶级实验室实验室所属机构知名教授方向SAIL(Stanford AI Lab)StanfordFei-Fei Li, Leonidas GuibasCV MLCSAILMITBill Freeman, Antonio TorralbaCV GraphicsBAIRUC BerkeleyAlyosha Efros, Trevor DarrellCV NLPVGG / OIIOxfordAndrew Zisserman, Andrea Vedaldi识别/视频INRIA (Willow)INRIA / ENSJean Ponce, Josef Sivic3D视觉MPI InformaticsMax PlanckBernt Schiele, Christian Theobalt人体/3DFacebook AI Research (FAIR)MetaYann LeCun, Michael Tschannen自监督/基础模型Google DeepMindGoogleNando de Freitas多模态/RLCVLETH ZurichLuc Van Gool3D/自动驾驶CMU RICMUTakeo Kanade, Martial Hebert机器人视觉商汤研究院SenseTime汤晓鸥全栈CV旷视研究院Megvii孙剑检测识别腾讯优图Tencent贾佳亚应用CV阿里巴巴达摩院Alibaba团队视觉AI百度视觉Baidu团队自动驾驶/OCR GitHub Awesome 系列仓库链接内容Awesome Computer Visionhttps://github.com/jbhuang0604/awesome-computer-vision综合CV资源列表Awesome Deep Visionhttps://github.com/kjw0612/awesome-deep-vision深度学习CVAwesome Object Detectionhttps://github.com/amusi/awesome-object-detection目标检测论文/代码Awesome 3D Reconstructionhttps://github.com/openMVG/awesome_3DReconstruction3D重建Awesome SLAMSLAM相关同步定位与建图Awesome NeRFhttps://github.com/awesome-NeRF/awesome-NeRFNeRF资源大全Awesome Vision TransformerVision Transformer相关ViT论文/代码Awesome Face Recognition人脸识别相关人脸识别文献Awesome Self-Supervised Learninghttps://github.com/jason718/awesome-self-supervised-learning自监督学习Awesome Diffusion Modelshttps://github.com/hee9joon/Awesome-Diffusion-Models扩散模型Papers With Codehttps://paperswithcode.com/论文代码BenchmarkSOTA Modelshttps://paperswithcode.com/sota各任务SOTA排行榜 关键数据集图像分类数据集图片数类别数用途链接ImageNet-1K1.2M1000通用分类ImageNetCIFAR-10/10060K10/100小规模分类https://www.cs.toronto.edu/~kriz/cifar.htmlMNIST70K10手写数字Index of /exdb/mnistFashion-MNIST70K10时尚分类GitHub - zalandoresearch/fashion-mnist: A MNIST-like fashion product database. Benchmark · GitHubPlaces3651.8M365场景分类http://places2.csail.mit.edu/目标检测数据集图片数类别数标注方式链接COCO330K80bboxsegCOCO - Common Objects in ContextPASCAL VOC11.5K20bboxsegThe PASCAL Visual Object Classes HomepageOpen Images9M600bboxseghttps://storage.googleapis.com/openimages/web/index.htmlVisDrone86K10无人机视角https://github.com/VisDrone/VisDrone-DatasetDOTA280K15旋转框https://captain-whu.github.io/DOTA/图像分割数据集任务链接ADE20K语义分割https://groups.csail.mit.edu/vision/datasets/ADE20K/Cityscapes街景分割https://www.cityscapes-dataset.com/Pascal Context场景理解http://www.cs.stanford.edu/~roozbeh/pascal-context/KITTI自动驾驶The KITTI Vision Benchmark SuiteCOCO-Stuff语义分割GitHub - nightrome/cocostuff: The official homepage of the COCO-Stuff dataset. · GitHub人脸数据集数量链接LFW13Khttp://vis-www.cs.umass.edu/lfw/MS-Celeb-1M10MMS-Celeb-1M: Challenge of Recognizing One Million Celebrities in the Real World - Microsoft ResearchVGGFace23.3MVisual Geometry Group - University of OxfordMegaFace1MMegaFace3D视觉数据集用途链接ShapeNet3D形状ShapeNetModelNet3D分类Princeton ModelNetScanNet室内3DScanNet | Richly-annotated 3D Reconstructions of Indoor ScenesKITTI自动驾驶3DThe KITTI Vision Benchmark SuiteWaymo Open自动驾驶https://waymo.com/open/nuScenes自动驾驶https://www.nuscenes.org/MegaDepth3D重建https://www.cs.cornell.edu/projects/megadepth/ 开发工具与框架核心框架PyTorch: https://pytorch.org/TensorFlow: https://www.tensorflow.org/JAX: GitHub - jax-ml/jax: Composable transformations of PythonNumPy programs: differentiate, vectorize, JIT to GPU/TPU, and more · GitHubKeras: Keras: Deep Learning for humansCV专用框架框架链接特点OpenCVOpenCV - Open Computer Vision Library经典CV库MMDetectionGitHub - open-mmlab/mmdetection: OpenMMLab Detection Toolbox and Benchmark · GitHub检测统一框架MMSegmentationGitHub - open-mmlab/mmsegmentation: OpenMMLab Semantic Segmentation Toolbox and Benchmark. · GitHub分割统一框架MMPoseGitHub - open-mmlab/mmpose: OpenMMLab Pose Estimation Toolbox and Benchmark. · GitHub姿态估计MMOCRhttps://github.com/open-mmlab/mmocrOCR工具箱MMTrackinghttps://github.com/open-mmlab/mmtracking目标跟踪Detectron2https://github.com/facebookresearch/detectron2Meta检测框架Segmentation Modelshttps://github.com/qubvel-org/segmentation_models.pytorchPyTorch分割TIMMhttps://github.com/huggingface/pytorch-image-models预训练模型库PyTorch3DGitHub - facebookresearch/pytorch3d: PyTorch3D is FAIRs library of reusable components for deep learning with 3D data · GitHub3D深度学习KorniaGitHub - kornia/kornia: Geometric Computer Vision Library for Spatial AI · GitHub可微分CV库FiftyOneGitHub - voxel51/fiftyone: Refine high-quality datasets and visual AI models · GitHub数据集可视化工具部署工具工具链接用途ONNXONNX | Home模型中间表示TensorRTTensorRT SDK | NVIDIA DeveloperNVIDIA推理优化OpenVINOhttps://docs.openvino.ai/Intel推理引擎TFLitehttps://www.tensorflow.org/lite移动端部署NCNNGitHub - Tencent/ncnn: ncnn is a high-performance neural network inference framework optimized for the mobile platform · GitHub腾讯移动端推理MNNGitHub - alibaba/MNN: MNN: A blazing-fast, lightweight inference engine battle-tested by Alibaba, powering high-performance on-device LLMs and Edge AI. · GitHub阿里巴巴推理框架TVMApache TVM端到端编译栈 学习平台与社区平台链接特点Papers With Codehttps://paperswithcode.com/论文代码排行榜Hugging Facehttps://huggingface.co/models模型库数据集SpacesKagglehttps://www.kaggle.com/CV竞赛数据集NotebookGitHubhttps://github.com/源代码复现Google Colabhttps://colab.research.google.com/免费GPURoboflowhttps://roboflow.com/CV数据管理与微调CVF开放访问https://openaccess.thecvf.com/CVPR/ICCV全文arXivhttps://arxiv.org/list/cs.CV/recent最新CV论文Google Scholarhttps://scholar.google.com/引文搜索知乎https://www.zhihu.com/topic/19551392中文CV讨论 论文下载与搜索工具链接arXiv (cs.CV)https://arxiv.org/list/cs.CV/recentarXiv Sanityhttps://arxiv-sanity-lite.com/ (Andrej Karpathy)Papers With Codehttps://paperswithcode.com/CVF Open Accesshttps://openaccess.thecvf.com/Semantic ScholarSemantic Scholar | AI-Powered Research ToolGoogle Scholarhttps://scholar.google.com/Connected Papershttps://www.connectedpapers.com/ResearchRabbitResearchRabbit: AI Tool for Smarter, Faster Literature ReviewsSci-Hub非必要不使用Aminerhttps://www.aminer.cn/ 2024-2026 前沿趋势方向代表性工作趋势说明视觉基础模型SAM 2, DINOv2, ViT-22B大一统模型few-shot/zero-shot多模态大模型GPT-4V, Gemini, LLaVA, Qwen-VL视觉语言融合3D生成3D Gaussian Splatting, Zero-1-to-33D AIGC爆发视频生成Sora, VideoPoet, Stable Video文生视频/图生视频可微渲染NeRF, 3DGS, Flash3D隐式/显式3D表示自监督DINOv2, I-JEPA, MAE减少对标注的依赖扩散模型SD3, Flux, DDPM图像生成主流范式世界模型UniSim, DayDreamer视觉物理仿真平铺CNN/ConvNeXtConvNeXt V2CNN复兴媲美ViT端侧部署MobileNetV4, EfficientViT手机上跑大模型对抗鲁棒性PGD-AT, TRADES, Random Smoothing让CV模型更安全可解释性Grad-CAM, LIME, SHAP, Attention Rollout理解模型决策原因 最后更新: 2026-06-17新增对抗攻击与防御、可解释性模块附录深层补充一、2024-2026 最新顶会论文获取渠道1.1 顶会论文搜索技巧渠道网址使用技巧CVF Open Accesshttps://openaccess.thecvf.com统一搜索 CVPR/ICCV/ECCV/WACV 全部论文全文Dblphttps://dblp.org按会议年份过滤如dblp.org/db/conf/cvpr/cvpr2025.htmlarXiv cs.CVComputer Vision and Pattern RecognitionRSS 订阅 按更新时间追踪最新论文Semantic Scholarhttps://www.semanticscholar.orgAPI 批量搜索结构化参考文献Google Scholarhttps://scholar.google.com引文追踪作者主页设置提醒新论文自动邮件通知1.2 Papers with Code 的高级筛选Papers with Code 不仅仅是论文代码的配对还提供强大的筛选功能按任务筛选paperswithcode.com/task/object-detection按不同 benchmark 查看 SOTA每个 benchmark 有 leaderboard 展示 Top 10 方法和数值可查看历史 SOTA 变迁按时间线的精度提升曲线按数据集筛选paperswithcode.com/datasets查看数据集页面每个数据集展示其相关的 benchmark 和 SOTA 方法按方法筛选支持搜索特定方法如 DETR和对比不同方法的 benchmark 表现RSS / Watchlist订阅特定任务/数据集新 SOTA 出现时自动通知1.3 Hugging Face Daily Papers每日前 5-10 篇最受关注的 arXiv 论文由社区共同投票upvote决定每天更新覆盖 CV / NLP / 多模态 / 语音 / 强化学习配合 Hugging Face 的 model card 可直接试验链接https://huggingface.co/papers二、实用的研究工具2.1 Connected Papers论文关系图Connected Papers | Find and explore academic papers输入一篇论文 → 生成论文影响力图谱节点论文越大代表引用越多或相关性越高边引用关系共被引、相关引用的强引用关系颜色按时间越深色越新功能Prior Works该论文基于的早期工作 Derivative Works后续发展用途快速了解一个领域的边界和关键路径发现意料之外的跨领域联系为综述提供系统化的文献框架2.2 Semantic Scholar研究搜索https://www.semanticscholar.org/AI 驱动的学术搜索引擎TLDRToo Long; Didnt ReadAI 生成的一句话论文摘要Citation Graph引用网络的结构化展示Influential Citations标记出最有影响力的引用API支持批量论文搜索和元数据获取Research Feed基于你的阅读历史推荐新论文2.3 Arxiv Sanity论文推荐https://arxiv-sanity-lite.com/Andrej Karpathy 维护从 arXiv 最新论文中根据你的偏好推荐通过点赞和收藏学习你的兴趣每周邮件摘要比直接刷 arXiv 高效得多2.4 科研 AI 助手工具功能链接Consensus检索一句话的科学答案附带引用https://consensus.appGalactica开源面向科研的 LLM支持公式/代码https://galactica.orgElicit自动化文献综述搜索和对比https://elicit.comSciSpace论文解读助手解释公式/图表https://typeset.ioPerplexity Pro实时搜索合成科研答案https://perplexity.ai三、中文社区资源3.1 知乎 CV 话题知乎专栏/话题描述链接计算机视觉话题30 万关注核心讨论区zhihu.com/topic/19551392AI 漫游指北李沐的知乎专栏高质量论文解读zhihu.com/column/c_1251902395475275776CV 前沿论文速递每日/每周最新论文zhihu.com/search?qCVPR2025论文计算机视觉与深度学习黄浴老师的技术分析知乎搜索黄浴3.2 公众号推荐公众号特点机器之心每日 AI 新闻覆盖全面量子位产业技术趋势分析CVerCV 入门到竞赛中文教程最全OpenMMLabOpenMMLab 官方算法开源详解AI 科技评论业界深度访谈Datawhale公益学习组织组团学 CV极市平台算法竞赛、工业应用深度学习初学者NewBeeNLP侧重 CV/DL 学习路径3.3 国内技术博客博客地址OpenMMLab 技术博客https://github.com/open-mmlab/OpenMMLabBlogDatawhale 教程Datawhale · GitHub动手学深度学习李沐https://d2l.ai/中文版CS231n 中文笔记ShowMeAI知识社区GiantPandaCV 公众号知乎/GitHub 搜 GiantPandaCVB 站 CV 论文精讲搜索论文精讲 计算机视觉3.4 国内竞赛平台平台描述链接天池阿里CV 比赛 数据集 GPU 算力https://tianchi.aliyun.comKaggle CN百度Kaggle 中文版https://kaggle.com中文社区Biendata国内数据竞赛https://www.biendata.xyzDataFountain数据竞赛https://www.datafountain.cn
