[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"similar-osmr--imgclsmob":3,"tool-osmr--imgclsmob":61},[4,18,26,36,44,52],{"id":5,"name":6,"github_repo":7,"description_zh":8,"stars":9,"difficulty_score":10,"last_commit_at":11,"category_tags":12,"status":17},4358,"openclaw","openclaw\u002Fopenclaw","OpenClaw 是一款专为个人打造的本地化 AI 助手，旨在让你在自己的设备上拥有完全可控的智能伙伴。它打破了传统 AI 助手局限于特定网页或应用的束缚，能够直接接入你日常使用的各类通讯渠道，包括微信、WhatsApp、Telegram、Discord、iMessage 等数十种平台。无论你在哪个聊天软件中发送消息，OpenClaw 都能即时响应，甚至支持在 macOS、iOS 和 Android 设备上进行语音交互，并提供实时的画布渲染功能供你操控。\n\n这款工具主要解决了用户对数据隐私、响应速度以及“始终在线”体验的需求。通过将 AI 部署在本地，用户无需依赖云端服务即可享受快速、私密的智能辅助，真正实现了“你的数据，你做主”。其独特的技术亮点在于强大的网关架构，将控制平面与核心助手分离，确保跨平台通信的流畅性与扩展性。\n\nOpenClaw 非常适合希望构建个性化工作流的技术爱好者、开发者，以及注重隐私保护且不愿被单一生态绑定的普通用户。只要具备基础的终端操作能力（支持 macOS、Linux 及 Windows WSL2），即可通过简单的命令行引导完成部署。如果你渴望拥有一个懂你",349277,3,"2026-04-06T06:32:30",[13,14,15,16],"Agent","开发框架","图像","数据工具","ready",{"id":19,"name":20,"github_repo":21,"description_zh":22,"stars":23,"difficulty_score":10,"last_commit_at":24,"category_tags":25,"status":17},3808,"stable-diffusion-webui","AUTOMATIC1111\u002Fstable-diffusion-webui","stable-diffusion-webui 是一个基于 Gradio 构建的网页版操作界面，旨在让用户能够轻松地在本地运行和使用强大的 Stable Diffusion 图像生成模型。它解决了原始模型依赖命令行、操作门槛高且功能分散的痛点，将复杂的 AI 绘图流程整合进一个直观易用的图形化平台。\n\n无论是希望快速上手的普通创作者、需要精细控制画面细节的设计师，还是想要深入探索模型潜力的开发者与研究人员，都能从中获益。其核心亮点在于极高的功能丰富度：不仅支持文生图、图生图、局部重绘（Inpainting）和外绘（Outpainting）等基础模式，还独创了注意力机制调整、提示词矩阵、负向提示词以及“高清修复”等高级功能。此外，它内置了 GFPGAN 和 CodeFormer 等人脸修复工具，支持多种神经网络放大算法，并允许用户通过插件系统无限扩展能力。即使是显存有限的设备，stable-diffusion-webui 也提供了相应的优化选项，让高质量的 AI 艺术创作变得触手可及。",162132,"2026-04-05T11:01:52",[14,15,13],{"id":27,"name":28,"github_repo":29,"description_zh":30,"stars":31,"difficulty_score":32,"last_commit_at":33,"category_tags":34,"status":17},1381,"everything-claude-code","affaan-m\u002Feverything-claude-code","everything-claude-code 是一套专为 AI 编程助手（如 Claude Code、Codex、Cursor 等）打造的高性能优化系统。它不仅仅是一组配置文件，而是一个经过长期实战打磨的完整框架，旨在解决 AI 代理在实际开发中面临的效率低下、记忆丢失、安全隐患及缺乏持续学习能力等核心痛点。\n\n通过引入技能模块化、直觉增强、记忆持久化机制以及内置的安全扫描功能，everything-claude-code 能显著提升 AI 在复杂任务中的表现，帮助开发者构建更稳定、更智能的生产级 AI 代理。其独特的“研究优先”开发理念和针对 Token 消耗的优化策略，使得模型响应更快、成本更低，同时有效防御潜在的攻击向量。\n\n这套工具特别适合软件开发者、AI 研究人员以及希望深度定制 AI 工作流的技术团队使用。无论您是在构建大型代码库，还是需要 AI 协助进行安全审计与自动化测试，everything-claude-code 都能提供强大的底层支持。作为一个曾荣获 Anthropic 黑客大奖的开源项目，它融合了多语言支持与丰富的实战钩子（hooks），让 AI 真正成长为懂上",141543,2,"2026-04-06T11:32:54",[14,13,35],"语言模型",{"id":37,"name":38,"github_repo":39,"description_zh":40,"stars":41,"difficulty_score":32,"last_commit_at":42,"category_tags":43,"status":17},2271,"ComfyUI","Comfy-Org\u002FComfyUI","ComfyUI 是一款功能强大且高度模块化的视觉 AI 引擎，专为设计和执行复杂的 Stable Diffusion 图像生成流程而打造。它摒弃了传统的代码编写模式，采用直观的节点式流程图界面，让用户通过连接不同的功能模块即可构建个性化的生成管线。\n\n这一设计巧妙解决了高级 AI 绘图工作流配置复杂、灵活性不足的痛点。用户无需具备编程背景，也能自由组合模型、调整参数并实时预览效果，轻松实现从基础文生图到多步骤高清修复等各类复杂任务。ComfyUI 拥有极佳的兼容性，不仅支持 Windows、macOS 和 Linux 全平台，还广泛适配 NVIDIA、AMD、Intel 及苹果 Silicon 等多种硬件架构，并率先支持 SDXL、Flux、SD3 等前沿模型。\n\n无论是希望深入探索算法潜力的研究人员和开发者，还是追求极致创作自由度的设计师与资深 AI 绘画爱好者，ComfyUI 都能提供强大的支持。其独特的模块化架构允许社区不断扩展新功能，使其成为当前最灵活、生态最丰富的开源扩散模型工具之一，帮助用户将创意高效转化为现实。",107888,"2026-04-06T11:32:50",[14,15,13],{"id":45,"name":46,"github_repo":47,"description_zh":48,"stars":49,"difficulty_score":10,"last_commit_at":50,"category_tags":51,"status":17},4487,"LLMs-from-scratch","rasbt\u002FLLMs-from-scratch","LLMs-from-scratch 是一个基于 PyTorch 的开源教育项目，旨在引导用户从零开始一步步构建一个类似 ChatGPT 的大型语言模型（LLM）。它不仅是同名技术著作的官方代码库，更提供了一套完整的实践方案，涵盖模型开发、预训练及微调的全过程。\n\n该项目主要解决了大模型领域“黑盒化”的学习痛点。许多开发者虽能调用现成模型，却难以深入理解其内部架构与训练机制。通过亲手编写每一行核心代码，用户能够透彻掌握 Transformer 架构、注意力机制等关键原理，从而真正理解大模型是如何“思考”的。此外，项目还包含了加载大型预训练权重进行微调的代码，帮助用户将理论知识延伸至实际应用。\n\nLLMs-from-scratch 特别适合希望深入底层原理的 AI 开发者、研究人员以及计算机专业的学生。对于不满足于仅使用 API，而是渴望探究模型构建细节的技术人员而言，这是极佳的学习资源。其独特的技术亮点在于“循序渐进”的教学设计：将复杂的系统工程拆解为清晰的步骤，配合详细的图表与示例，让构建一个虽小但功能完备的大模型变得触手可及。无论你是想夯实理论基础，还是为未来研发更大规模的模型做准备",90106,"2026-04-06T11:19:32",[35,15,13,14],{"id":53,"name":54,"github_repo":55,"description_zh":56,"stars":57,"difficulty_score":10,"last_commit_at":58,"category_tags":59,"status":17},4292,"Deep-Live-Cam","hacksider\u002FDeep-Live-Cam","Deep-Live-Cam 是一款专注于实时换脸与视频生成的开源工具，用户仅需一张静态照片，即可通过“一键操作”实现摄像头画面的即时变脸或制作深度伪造视频。它有效解决了传统换脸技术流程繁琐、对硬件配置要求极高以及难以实时预览的痛点，让高质量的数字内容创作变得触手可及。\n\n这款工具不仅适合开发者和技术研究人员探索算法边界，更因其极简的操作逻辑（仅需三步：选脸、选摄像头、启动），广泛适用于普通用户、内容创作者、设计师及直播主播。无论是为了动画角色定制、服装展示模特替换，还是制作趣味短视频和直播互动，Deep-Live-Cam 都能提供流畅的支持。\n\n其核心技术亮点在于强大的实时处理能力，支持口型遮罩（Mouth Mask）以保留使用者原始的嘴部动作，确保表情自然精准；同时具备“人脸映射”功能，可同时对画面中的多个主体应用不同面孔。此外，项目内置了严格的内容安全过滤机制，自动拦截涉及裸露、暴力等不当素材，并倡导用户在获得授权及明确标注的前提下合规使用，体现了技术发展与伦理责任的平衡。",88924,"2026-04-06T03:28:53",[14,15,13,60],"视频",{"id":62,"github_repo":63,"name":64,"description_en":65,"description_zh":66,"ai_summary_zh":67,"readme_en":68,"readme_zh":69,"quickstart_zh":70,"use_case_zh":71,"hero_image_url":72,"owner_login":73,"owner_name":74,"owner_avatar_url":75,"owner_bio":76,"owner_company":77,"owner_location":77,"owner_email":78,"owner_twitter":77,"owner_website":77,"owner_url":79,"languages":80,"stars":92,"forks":93,"last_commit_at":94,"license":95,"difficulty_score":32,"env_os":96,"env_gpu":97,"env_ram":96,"env_deps":98,"category_tags":107,"github_topics":109,"view_count":32,"oss_zip_url":77,"oss_zip_packed_at":77,"status":17,"created_at":129,"updated_at":130,"faqs":131,"releases":162},4511,"osmr\u002Fimgclsmob","imgclsmob","Sandbox for training deep learning networks","imgclsmob 是一个专注于计算机视觉任务的深度学习网络研究沙盒。它致力于解决研究人员和开发者在复现经典模型、跨框架对比实验以及获取高质量预训练权重时面临的繁琐问题。通过提供统一的接口，imgclsmob 集成了分类、分割、目标检测和姿态估计等多种主流模型的实现与训练脚本。\n\n其最显著的技术亮点在于强大的多框架支持能力。无论是 MXNet\u002FGluon、PyTorch、Chainer、Keras，还是 TensorFlow 1.x\u002F2.x，imgclsmob 均提供了对应的纯模型包（如 pytorchcv、tf2cv 等），且大部分模型优先在 Gluon 上实现后迁移至其他平台，确保了算法的一致性。此外，项目内置了基于 ImageNet、CIFAR、COCO 等权威数据集的预训练权重，支持在调用时自动加载，极大降低了实验门槛。\n\nimgclsmob 非常适合 AI 研究人员、算法工程师及深度学习爱好者使用。对于需要快速验证新想法、进行模型性能基准测试，或希望在不同深度学习框架间无缝切换的用户而言，它是一个高效、可靠且开源的基础设施工具，能帮助用户将精力更集中于核心算法研究而非重复性的","imgclsmob 是一个专注于计算机视觉任务的深度学习网络研究沙盒。它致力于解决研究人员和开发者在复现经典模型、跨框架对比实验以及获取高质量预训练权重时面临的繁琐问题。通过提供统一的接口，imgclsmob 集成了分类、分割、目标检测和姿态估计等多种主流模型的实现与训练脚本。\n\n其最显著的技术亮点在于强大的多框架支持能力。无论是 MXNet\u002FGluon、PyTorch、Chainer、Keras，还是 TensorFlow 1.x\u002F2.x，imgclsmob 均提供了对应的纯模型包（如 pytorchcv、tf2cv 等），且大部分模型优先在 Gluon 上实现后迁移至其他平台，确保了算法的一致性。此外，项目内置了基于 ImageNet、CIFAR、COCO 等权威数据集的预训练权重，支持在调用时自动加载，极大降低了实验门槛。\n\nimgclsmob 非常适合 AI 研究人员、算法工程师及深度学习爱好者使用。对于需要快速验证新想法、进行模型性能基准测试，或希望在不同深度学习框架间无缝切换的用户而言，它是一个高效、可靠且开源的基础设施工具，能帮助用户将精力更集中于核心算法研究而非重复性的工程实现。","# Deep learning networks\n\n[![Build Status](https:\u002F\u002Ftravis-ci.org\u002Fosmr\u002Fimgclsmob.svg?branch=master)](https:\u002F\u002Ftravis-ci.org\u002Fosmr\u002Fimgclsmob)\n[![GitHub License](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FLicense-MIT-blue.svg)](https:\u002F\u002Fopensource.org\u002Flicenses\u002FMIT)\n[![Python Version](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fpython-3.10-lightgrey.svg)](https:\u002F\u002Fgithub.com\u002Fosmr\u002Fimgclsmob)\n\nThis repo is used to research convolutional networks primarily for computer vision tasks. For this purpose, the repo\ncontains (re)implementations of various classification, segmentation, detection, and pose estimation models and scripts\nfor training\u002Fevaluating\u002Fconverting.\n\nThe following frameworks are used:\n- MXNet\u002FGluon ([info](https:\u002F\u002Fmxnet.apache.org)),\n- PyTorch ([info](https:\u002F\u002Fpytorch.org)),\n- Chainer ([info](https:\u002F\u002Fchainer.org)),\n- Keras ([info](https:\u002F\u002Fkeras.io)),\n- TensorFlow 1.x\u002F2.x ([info](https:\u002F\u002Fwww.tensorflow.org)).\n\nFor each supported framework, there is a PIP-package containing pure models without auxiliary scripts. List of packages:\n- [gluoncv2](https:\u002F\u002Fpypi.org\u002Fproject\u002Fgluoncv2) for Gluon,\n- [pytorchcv](https:\u002F\u002Fpypi.org\u002Fproject\u002Fpytorchcv) for PyTorch,\n- [chainercv2](https:\u002F\u002Fpypi.org\u002Fproject\u002Fchainercv2) for Chainer,\n- [kerascv](https:\u002F\u002Fpypi.org\u002Fproject\u002Fkerascv) for Keras,\n- [tensorflowcv](https:\u002F\u002Fpypi.org\u002Fproject\u002Ftensorflowcv) for TensorFlow 1.x,\n- [tf2cv](https:\u002F\u002Fpypi.org\u002Fproject\u002Ftf2cv) for TensorFlow 2.x.\n\nCurrently, models are mostly implemented on Gluon and then ported to other frameworks. Some models are pretrained on\n[ImageNet-1K](http:\u002F\u002Fwww.image-net.org), [CIFAR-10\u002F100](https:\u002F\u002Fwww.cs.toronto.edu\u002F~kriz\u002Fcifar.html),\n[SVHN](http:\u002F\u002Fufldl.stanford.edu\u002Fhousenumbers), [CUB-200-2011](http:\u002F\u002Fwww.vision.caltech.edu\u002Fvisipedia\u002FCUB-200-2011.html),\n[Pascal VOC2012](http:\u002F\u002Fhost.robots.ox.ac.uk\u002Fpascal\u002FVOC\u002Fvoc2012), [ADE20K](http:\u002F\u002Fgroups.csail.mit.edu\u002Fvision\u002Fdatasets\u002FADE20K),\n[Cityscapes](https:\u002F\u002Fwww.cityscapes-dataset.com), and [COCO](http:\u002F\u002Fcocodataset.org) datasets. All pretrained weights\nare loaded automatically during use. See examples of such automatic loading of weights in the corresponding sections of\nthe documentation dedicated to a particular package:\n- [Gluon models](gluon\u002FREADME.md),\n- [PyTorch models](https:\u002F\u002Fgithub.com\u002Fosmr\u002Fpytorchcv\u002Fblob\u002Fmaster\u002FREADME.md),\n- [Chainer models](chainer_\u002FREADME.md),\n- [Keras models](keras_\u002FREADME.md),\n- [TensorFlow 1.x models](tensorflow_\u002FREADME.md),\n- [TensorFlow 2.x models](tensorflow2\u002FREADME.md).\n\n## Installation\n\nTo use training\u002Fevaluating scripts as well as all models, you need to clone the repository and install dependencies:\n```\ngit clone git@github.com:osmr\u002Fimgclsmob.git\npip install -r requirements.txt\n```\n\n## Table of implemented classification models\n\nSome remarks:\n- `Repo` is an author repository, if it exists.\n- `a`, `b`, `c`, `d`, and `e` means the implementation of a model for ImageNet-1K, CIFAR-10, CIFAR-100, SVHN, and CUB-200-2011, respectively.\n- `A`, `B`, `C`, `D`, and `E` means having a pre-trained model for corresponding datasets.\n\n| Model | [Gluon](gluon\u002FREADME.md) | [PyTorch](https:\u002F\u002Fgithub.com\u002Fosmr\u002Fpytorchcv\u002Fblob\u002Fmaster\u002FREADME.md) | [Chainer](chainer_\u002FREADME.md) | [Keras](keras_\u002FREADME.md) | [TF](tensorflow_\u002FREADME.md) | [TF2](tensorflow2\u002FREADME.md) | Paper | Repo | Year |\n| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |\n| AlexNet | A | A | A | A | A | A | [link](http:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F4824-imagenet-classification-with-deep-convolutional-neural-networks.pdf) | [link](https:\u002F\u002Fcode.google.com\u002Farchive\u002Fp\u002Fcuda-convnet2) | 2012 |\n| ZFNet | A | A | A | A | A | A | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1311.2901) | - | 2013 |\n| VGG | A | A | A | A | A | A | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1409.1556) | - | 2014 |\n| BN-VGG | A | A | A | A | A | A | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1409.1556) | - | 2015 |\n| BN-Inception | A | A | A | - | - | A | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1502.03167) | - | 2015 |\n| ResNet | ABCDE | ABCDE | ABCDE | A | A | ABCDE | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1512.03385) | [link](https:\u002F\u002Fgithub.com\u002FKaimingHe\u002Fdeep-residual-networks) | 2015 |\n| PreResNet | ABCD | ABCD | ABCD | A | A | ABCD | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1603.05027) | [link](https:\u002F\u002Fgithub.com\u002Ffacebook\u002Ffb.resnet.torch) | 2016 |\n| ResNeXt | ABCD | ABCD | ABCD | A | A | ABCD | [link](http:\u002F\u002Farxiv.org\u002Fabs\u002F1611.05431) | [link](https:\u002F\u002Fgithub.com\u002Ffacebookresearch\u002FResNeXt) | 2016 |\n| SENet | A | A | A | A | A | A | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1709.01507) | [link](https:\u002F\u002Fgithub.com\u002Fhujie-frank\u002FSENet) | 2017 |\n| SE-ResNet | ABCDE | ABCDE | ABCDE | A | A | ABCDE | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1709.01507) | [link](https:\u002F\u002Fgithub.com\u002Fhujie-frank\u002FSENet) | 2017 |\n| SE-PreResNet | ABCD | ABCD | ABCD | A | A | ABCD | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1709.01507) | [link](https:\u002F\u002Fgithub.com\u002Fhujie-frank\u002FSENet) | 2017 |\n| SE-ResNeXt | A | A | A | A | A | A | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1709.01507) | [link](https:\u002F\u002Fgithub.com\u002Fhujie-frank\u002FSENet) | 2017 |\n| ResNeSt(A) | A | A | A | - | - | A | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F2004.08955) | [link](https:\u002F\u002Fgithub.com\u002Fzhanghang1989\u002FResNeSt) | 2020 |\n| IBN-ResNet | A | A | - | - | - | A | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1807.09441) | [link](https:\u002F\u002Fgithub.com\u002FXingangPan\u002FIBN-Net) | 2018 |\n| IBN-ResNeXt | A | A | - | - | - | A | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1807.09441) | [link](https:\u002F\u002Fgithub.com\u002FXingangPan\u002FIBN-Net) | 2018 |\n| IBN-DenseNet | A | A | - | - | - | A | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1807.09441) | [link](https:\u002F\u002Fgithub.com\u002FXingangPan\u002FIBN-Net) | 2018 |\n| AirNet | A | A | A | - | - | A | [link](https:\u002F\u002Fieeexplore.ieee.org\u002Fdocument\u002F8510896) | [link](https:\u002F\u002Fgithub.com\u002Fsoeaver\u002FAirNet-PyTorch) | 2018 |\n| AirNeXt | A | A | A | - | - | A | [link](https:\u002F\u002Fieeexplore.ieee.org\u002Fdocument\u002F8510896) | [link](https:\u002F\u002Fgithub.com\u002Fsoeaver\u002FAirNet-PyTorch) | 2018 |\n| BAM-ResNet | A | A | A | - | - | A | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1807.06514) | [link](https:\u002F\u002Fgithub.com\u002FJongchan\u002Fattention-module) | 2018 |\n| CBAM-ResNet | A | A | A | - | - | A | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1807.06521) | [link](https:\u002F\u002Fgithub.com\u002FJongchan\u002Fattention-module) | 2018 |\n| ResAttNet | a | a | a | - | - | - | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1704.06904) | [link](https:\u002F\u002Fgithub.com\u002Ffwang91\u002Fresidual-attention-network) | 2017 |\n| SKNet | a | a | a | - | - | - | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1903.06586) | [link](https:\u002F\u002Fgithub.com\u002Fimplus\u002FSKNet) | 2019 |\n| SCNet | A | A | A | - | - | A | [link](http:\u002F\u002Fmftp.mmcheng.net\u002FPapers\u002F20cvprSCNet.pdf) | [link](https:\u002F\u002Fgithub.com\u002FMCG-NKU\u002FSCNet) | 2020 |\n| RegNet | A | A | A | - | - | A | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F2003.13678) | [link](https:\u002F\u002Fgithub.com\u002Ffacebookresearch\u002Fpycls) | 2020 |\n| DIA-ResNet | aBCD | aBCD | aBCD | - | - | - | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1905.10671) | [link](https:\u002F\u002Fgithub.com\u002Fgbup-group\u002FDIANet) | 2019 |\n| DIA-PreResNet | aBCD | aBCD | aBCD | - | - | - | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1905.10671) | [link](https:\u002F\u002Fgithub.com\u002Fgbup-group\u002FDIANet) | 2019 |\n| PyramidNet | ABCD | ABCD | ABCD | - | - | ABCD | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1610.02915) | [link](https:\u002F\u002Fgithub.com\u002Fjhkim89\u002FPyramidNet) | 2016 |\n| DiracNetV2 | A | A | A | - | - | A | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1706.00388) | [link](https:\u002F\u002Fgithub.com\u002Fszagoruyko\u002Fdiracnets) | 2017 |\n| ShaResNet | a | a | a | - | - | - | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1702.08782) | [link](https:\u002F\u002Fgithub.com\u002Faboulch\u002Fsharesnet) | 2017 |\n| CRU-Net | A | - | - | - | - | - | [link](https:\u002F\u002Fwww.ijcai.org\u002Fproceedings\u002F2018\u002F88) | [link](https:\u002F\u002Fgithub.com\u002Fcypw\u002FCRU-Net) | 2018 |\n| DenseNet | ABCD | ABCD | ABCD | A | A | ABCD | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1608.06993) | [link](https:\u002F\u002Fgithub.com\u002Fliuzhuang13\u002FDenseNet) | 2016 |\n| CondenseNet | A | A | A | - | - | - | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1711.09224) | [link](https:\u002F\u002Fgithub.com\u002FShichenLiu\u002FCondenseNet) | 2017 |\n| SparseNet | a | a | a | - | - | - | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1801.05895) | [link](https:\u002F\u002Fgithub.com\u002FLyken17\u002FSparseNet) | 2018 |\n| PeleeNet | A | A | A | - | - | A | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1804.06882) | [link](https:\u002F\u002Fgithub.com\u002FRobert-JunWang\u002FPelee) | 2018 |\n| Oct-ResNet | abcd | a | a | - | - | - | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1904.05049) | - | 2019 |\n| Res2Net | a | - | - | - | - | - | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1904.01169) | - | 2019 |\n| WRN | ABCD | ABCD | ABCD | - | - | a | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1605.07146) | [link](https:\u002F\u002Fgithub.com\u002Fszagoruyko\u002Fwide-residual-networks) | 2016 |\n| WRN-1bit | BCD | BCD | BCD | - | - | - | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1802.08530) | [link](https:\u002F\u002Fgithub.com\u002FMcDonnell-Lab\u002F1-bit-per-weight) | 2018 |\n| DRN-C | A | A | A | - | - | A | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1705.09914) | [link](https:\u002F\u002Fgithub.com\u002Ffyu\u002Fdrn) | 2017 |\n| DRN-D | A | A | A | - | - | A | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1705.09914) | [link](https:\u002F\u002Fgithub.com\u002Ffyu\u002Fdrn) | 2017 |\n| DPN | A | A | A | - | - | A | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1707.01629) | [link](https:\u002F\u002Fgithub.com\u002Fcypw\u002FDPNs) | 2017 |\n| DarkNet Ref | A | A | A | A | A | A | [link](https:\u002F\u002Fgithub.com\u002Fpjreddie\u002Fdarknet) | [link](https:\u002F\u002Fgithub.com\u002Fpjreddie\u002Fdarknet) | - |\n| DarkNet Tiny | A | A | A | A | A | A | [link](https:\u002F\u002Fgithub.com\u002Fpjreddie\u002Fdarknet) | [link](https:\u002F\u002Fgithub.com\u002Fpjreddie\u002Fdarknet) | - |\n| DarkNet-19 | a | a | a | a | a | a | [link](https:\u002F\u002Fgithub.com\u002Fpjreddie\u002Fdarknet) | [link](https:\u002F\u002Fgithub.com\u002Fpjreddie\u002Fdarknet) | - |\n| DarkNet-53 | A | A | A | A | A | A | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1804.02767) | [link](https:\u002F\u002Fgithub.com\u002Fpjreddie\u002Fdarknet) | 2018 |\n| ChannelNet | a | a | a | - | a | - | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1809.01330) | [link](https:\u002F\u002Fgithub.com\u002FHongyangGao\u002FChannelNets) | 2018 |\n| iSQRT-COV-ResNet | a | a | - | - | - | - | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1712.01034) | [link](https:\u002F\u002Fgithub.com\u002Fjiangtaoxie\u002Ffast-MPN-COV) | 2017 |\n| RevNet | - | a | - | - | - | - | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1707.04585) | [link](https:\u002F\u002Fgithub.com\u002Frenmengye\u002Frevnet-public) | 2017 |\n| i-RevNet | A | A | A | - | - | - | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1802.07088) | [link](https:\u002F\u002Fgithub.com\u002Fjhjacobsen\u002Fpytorch-i-revnet) | 2018 |\n| BagNet | A | A | A | - | - | A | [link](https:\u002F\u002Fopenreview.net\u002Fpdf?id=SkfMWhAqYQ) | [link](https:\u002F\u002Fgithub.com\u002Fwielandbrendel\u002Fbag-of-local-features-models) | 2019 |\n| DLA | A | A | A | - | - | A | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1707.06484) | [link](https:\u002F\u002Fgithub.com\u002Fucbdrive\u002Fdla) | 2017 |\n| MSDNet | a | ab | - | - | - | - | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1703.09844) | [link](https:\u002F\u002Fgithub.com\u002Fgaohuang\u002FMSDNet) | 2017 |\n| FishNet | A | A | A | - | - | - | [link](http:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F7356-fishnet-a-versatile-backbone-for-image-region-and-pixel-level-prediction.pdf) | [link](https:\u002F\u002Fgithub.com\u002Fkevin-ssy\u002FFishNet) | 2018 |\n| ESPNetv2 | A | A | A | - | - | - | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1811.11431) | [link](https:\u002F\u002Fgithub.com\u002Fsacmehta\u002FESPNetv2) | 2018 |\n| DiCENet | A | A | A | - | - | A | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1906.03516) | [link](https:\u002F\u002Fgithub.com\u002Fsacmehta\u002FEdgeNets) | 2019 |\n| HRNet | A | A | A | - | - | A | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1908.07919) | [link](https:\u002F\u002Fgithub.com\u002FHRNet\u002FHRNet-Image-Classification) | 2019 |\n| VoVNet | A | A | A | - | - | A | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1904.09730) | [link](https:\u002F\u002Fgithub.com\u002Fstigma0617\u002FVoVNet.pytorch) | 2019 |\n| SelecSLS | A | A | A | - | - | A | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1907.00837) | [link](https:\u002F\u002Fgithub.com\u002Fmehtadushy\u002FSelecSLS-Pytorch) | 2019 |\n| HarDNet | A | A | A | - | - | A | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1909.00948) | [link](https:\u002F\u002Fgithub.com\u002FPingoLH\u002FPytorch-HarDNet) | 2019 |\n| X-DenseNet | aBCD | aBCD | aBCD | - | - | - | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1711.08757) | [link](https:\u002F\u002Fgithub.com\u002FDrImpossible\u002FDeep-Expander-Networks) | 2017 |\n| SqueezeNet | A | A | A | A | A | A | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1602.07360) | [link](https:\u002F\u002Fgithub.com\u002FDeepScale\u002FSqueezeNet) | 2016 |\n| SqueezeResNet | A | A | A | A | A | A | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1602.07360) | - | 2016 |\n| SqueezeNext | A | A | A | A | A | A | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1803.10615) | [link](https:\u002F\u002Fgithub.com\u002Famirgholami\u002FSqueezeNext) | 2018 |\n| ShuffleNet | A | A | A | A | A | A | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1707.01083) | - | 2017 |\n| ShuffleNetV2 | A | A | A | A | A | A | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1807.11164) | - | 2018 |\n| MENet | A | A | A | A | A | A | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1803.09127) | [link](https:\u002F\u002Fgithub.com\u002Fclavichord93\u002FMENet) | 2018 |\n| MobileNet | AE | AE | AE | A | A | AE | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1704.04861) | [link](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodels) | 2017 |\n| FD-MobileNet | A | A | A | A | A | A | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1802.03750) | [link](https:\u002F\u002Fgithub.com\u002Fclavichord93\u002FFD-MobileNet) | 2018 |\n| MobileNetV2 | A | A | A | A | A | A | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1801.04381) | [link](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodels) | 2018 |\n| MobileNetV3 | A | A | A | A | - | A | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1905.02244) | [link](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodels) | 2019 |\n| IGCV3 | A | A | A | A | A | A | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1806.00178) | [link](https:\u002F\u002Fgithub.com\u002Fhomles11\u002FIGCV3) | 2018 |\n| GhostNet | a | a | a | - | - | a | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1911.11907) | [link](https:\u002F\u002Fgithub.com\u002Fiamhankai\u002Fghostnet) | 2019 |\n| MnasNet | A | A | A | A | A | A | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1807.11626) | - | 2018 |\n| DARTS | A | A | A | - | - | - | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1806.09055) | [link](https:\u002F\u002Fgithub.com\u002Fquark0\u002Fdarts) | 2018 |\n| ProxylessNAS | AE | AE | AE | - | - | AE | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1812.00332) | [link](https:\u002F\u002Fgithub.com\u002Fmit-han-lab\u002FProxylessNAS) | 2018 |\n| FBNet-C | A | A | A | - | - | A | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1812.03443) | - | 2018 |\n| Xception | A | A | A | - | - | A | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1610.02357) | [link](https:\u002F\u002Fgithub.com\u002Ffchollet\u002Fdeep-learning-models) | 2016 |\n| InceptionV3 | A | A | A | - | - | A | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1512.00567) | [link](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodels) | 2015 |\n| InceptionV4 | A | A | A | - | - | A | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1602.07261) | [link](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodels) | 2016 |\n| InceptionResNetV1 | A | A | A | - | - | A | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1602.07261) | [link](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodels) | 2016 |\n| InceptionResNetV2 | A | A | A | - | - | A | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1602.07261) | [link](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodels) | 2016 |\n| PolyNet | A | A | A | - | - | A | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1611.05725) | [link](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fpolynet) | 2016 |\n| NASNet-Large | A | A | A | - | - | A | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1707.07012) | [link](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodels) | 2017 |\n| NASNet-Mobile | A | A | A | - | - | A | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1707.07012) | [link](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodels) | 2017 |\n| PNASNet-Large | A | A | A | - | - | A | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1712.00559) | [link](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodels) | 2017 |\n| SPNASNet | A | A | A | - | - | A | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1904.02877) | [link](https:\u002F\u002Fgithub.com\u002Fdstamoulis\u002Fsingle-path-nas) | 2019 |\n| EfficientNet | A | A | A | A | - | A | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1905.11946) | [link](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Ftpu\u002Ftree\u002Fmaster\u002Fmodels\u002Fofficial\u002Fefficientnet) | 2019 |\n| MixNet | A | A | A | - | - | A | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1907.09595) | [link](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Ftpu\u002Ftree\u002Fmaster\u002Fmodels\u002Fofficial\u002Fmnasnet\u002Fmixnet) | 2019 |\n| NIN | BCD | BCD | BCD | - | - | - | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1312.4400) | [link](https:\u002F\u002Fgist.github.com\u002Fmavenlin\u002Fe56253735ef32c3c296d) | 2013 |\n| RoR-3 | BCD | BCD | BCD | - | - | - | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1608.02908) | - | 2016 |\n| RiR | BCD | BCD | BCD | - | - | - | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1603.08029) | - | 2016 |\n| ResDrop-ResNet | bcd | bcd | bcd | - | - | - | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1603.09382) | [link](https:\u002F\u002Fgithub.com\u002Fyueatsprograms\u002FStochastic_Depth) | 2016 |\n| Shake-Shake-ResNet | BCD | BCD | BCD | - | - | - | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1705.07485) | [link](https:\u002F\u002Fgithub.com\u002Fxgastaldi\u002Fshake-shake) | 2017 |\n| ShakeDrop-ResNet | bcd | bcd | bcd | - | - | - | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1802.02375) | - | 2018 |\n| FractalNet | bc | bc | - | - | - | - | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1605.07648) | [link](https:\u002F\u002Fgithub.com\u002Fgustavla\u002Ffractalnet) | 2016 |\n| NTS-Net | E | E | E | - | - | - | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1809.00287) | [link](https:\u002F\u002Fgithub.com\u002Fyangze0930\u002FNTS-Net) | 2018 |\n\n## Table of implemented segmentation models\n\nSome remarks:\n- `a\u002FA` corresponds to Pascal VOC2012.\n- `b\u002FB` corresponds to ADE20K.\n- `c\u002FC` corresponds to Cityscapes.\n- `d\u002FD` corresponds to COCO.\n- `e\u002FE` corresponds to CelebAMask-HQ.\n\n| Model | [Gluon](gluon\u002FREADME.md) | [PyTorch](https:\u002F\u002Fgithub.com\u002Fosmr\u002Fpytorchcv\u002Fblob\u002Fmaster\u002FREADME.md) | [Chainer](chainer_\u002FREADME.md) | [Keras](keras_\u002FREADME.md) | [TF](tensorflow_\u002FREADME.md)  | [TF2](tensorflow_\u002FREADME.md) | Paper | Repo | Year |\n| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |\n| PSPNet | ABCD | ABCD | ABCD | - | - | ABCD | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1612.01105) | - | 2016 |\n| DeepLabv3 | ABcD | ABcD | ABcD | - | - | ABcD | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1706.05587) | - | 2017 |\n| FCN-8s(d) | ABcD | ABcD | ABcD | - | - | ABcD | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1411.4038) | - | 2014 |\n| ICNet | C | C | C | - | - | C | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1704.08545) | [link](https:\u002F\u002Fgithub.com\u002Fhszhao\u002FICNet) | 2017 |\n| SINet | C | C | C | - | - | c | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1911.09099) | [link](https:\u002F\u002Fgithub.com\u002Fclovaai\u002Fc3_sinet) | 2019 |\n| BiSeNet | e | e | e | - | - | e | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1808.00897) | - | 2018 |\n| DANet | C | C | C | - | - | C | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1809.02983) | [link](https:\u002F\u002Fgithub.com\u002Fjunfu1115\u002FDANet) | 2018 |\n| Fast-SCNN | C | C | C | - | - | C | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1902.04502) | - | 2019 |\n| CGNet | c | c | c | - | - | c | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1811.08201) | [link](https:\u002F\u002Fgithub.com\u002FwutianyiRosun\u002FCGNet) | 2018 |\n| DABNet | c | c | c | - | - | c | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1907.11357) | [link](https:\u002F\u002Fgithub.com\u002FReagan1311\u002FDABNet) | 2019 |\n| FPENet | c | c | c | - | - | c | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1909.08599) | - | 2019 |\n| ContextNet | - | c | - | - | - | - | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1805.04554) | - | 2018 |\n| LEDNet | c | c | c | - | - | c | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1905.02423) | - | 2019 |\n| ESNet | - | c | - | - | - | - | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1906.09826) | - | 2019 |\n| EDANet | - | c | - | - | - | - | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1809.06323) | [link](https:\u002F\u002Fgithub.com\u002Fshaoyuanlo\u002FEDANet) | 2018 |\n| ENet | - | c | - | - | - | - | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1606.02147) | - | 2016 |\n| ERFNet | - | c | - | - | - | - | [link](http:\u002F\u002Fwww.robesafe.uah.es\u002Fpersonal\u002Feduardo.romera\u002Fpdfs\u002FRomera17tits.pdf) | - | 2017 |\n| LinkNet | - | c | - | - | - | - | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1707.03718) | - | 2017 |\n| SegNet | - | c | - | - | - | - | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1511.00561) | - | 2015 |\n| U-Net | - | c | - | - | - | - | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1505.04597) | - | 2015 |\n| SQNet | - | c | - | - | - | - | [link](https:\u002F\u002Fopenreview.net\u002Fpdf?id=S1uHiFyyg) | - | 2016 |\n\n## Table of implemented object detection models\n\nSome remarks:\n- `a\u002FA` corresponds to COCO.\n\n| Model | [Gluon](gluon\u002FREADME.md) | [PyTorch](https:\u002F\u002Fgithub.com\u002Fosmr\u002Fpytorchcv\u002Fblob\u002Fmaster\u002FREADME.md) | [Chainer](chainer_\u002FREADME.md) | [Keras](keras_\u002FREADME.md) | [TF](tensorflow_\u002FREADME.md)  | [TF2](tensorflow2\u002FREADME.md) | Paper | Repo | Year |\n| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |\n| CenterNet | a | a | a | - | - | a | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1904.07850) | [link](https:\u002F\u002Fgithub.com\u002Fxingyizhou\u002FCenterNet) | 2019 |\n\n## Table of implemented human pose estimation models\n\nSome remarks:\n- `a\u002FA` corresponds to COCO.\n\n| Model | [Gluon](gluon\u002FREADME.md) | [PyTorch](https:\u002F\u002Fgithub.com\u002Fosmr\u002Fpytorchcv\u002Fblob\u002Fmaster\u002FREADME.md) | [Chainer](chainer_\u002FREADME.md) | [Keras](keras_\u002FREADME.md) | [TF](tensorflow_\u002FREADME.md)  | [TF2](tensorflow2\u002FREADME.md) | Paper | Repo | Year |\n| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |\n| AlphaPose | A | A | A | - | - | A | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1612.00137) | [link](https:\u002F\u002Fgithub.com\u002FMVIG-SJTU\u002FAlphaPose) | 2016 |\n| SimplePose | A | A | A | - | - | A | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1804.06208) | [link](https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Fhuman-pose-estimation.pytorch) | 2018 |\n| SimplePose(Mobile) | A | A | A | - | - | A | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1804.06208) | - | 2018 |\n| Lightweight OpenPose | A | A | A | - | - | A | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1811.12004) | [link](https:\u002F\u002Fgithub.com\u002FDaniil-Osokin\u002Flightweight-human-pose-estimation-3d-demo.pytorch) | 2018 |\n| IBPPose | A | A | A | - | - | A | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1911.10529) | [link](https:\u002F\u002Fgithub.com\u002Fjialee93\u002FImproved-Body-Parts) | 2019 |\n\n## Table of implemented automatic speech recognition models\n\nSome remarks:\n- `a\u002FA` corresponds to LibriSpeech.\n- `b\u002FB` corresponds to Mozilla Common Voice.\n\n| Model | [Gluon](gluon\u002FREADME.md) | [PyTorch](https:\u002F\u002Fgithub.com\u002Fosmr\u002Fpytorchcv\u002Fblob\u002Fmaster\u002FREADME.md) | [Chainer](chainer_\u002FREADME.md) | [Keras](keras_\u002FREADME.md) | [TF](tensorflow_\u002FREADME.md)  | [TF2](tensorflow2\u002FREADME.md) | Paper | Repo | Year |\n| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |\n| Jasper DR | AB | AB | ab | - | - | ab | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1904.03288) | [link](https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FNeMo) | 2019 |\n| QuartzNet | AB | AB | ab | - | - | ab | [link](https:\u002F\u002Farxiv.org\u002Fabs\u002F1910.10261) | [link](https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FNeMo) | 2019 |\n","# 深度学习网络\n\n[![构建状态](https:\u002F\u002Ftravis-ci.org\u002Fosmr\u002Fimgclsmob.svg?branch=master)](https:\u002F\u002Ftravis-ci.org\u002Fosmr\u002Fimgclsmob)\n[![GitHub 许可证](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FLicense-MIT-blue.svg)](https:\u002F\u002Fopensource.org\u002Flicenses\u002FMIT)\n[![Python 版本](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fpython-3.10-lightgrey.svg)](https:\u002F\u002Fgithub.com\u002Fosmr\u002Fimgclsmob)\n\n此仓库主要用于研究卷积神经网络，特别是针对计算机视觉任务。为此，仓库包含了多种分类、分割、检测和姿态估计模型的（重新）实现，以及用于训练、评估和转换的脚本。\n\n所使用的框架包括：\n- MXNet\u002FGluon ([信息](https:\u002F\u002Fmxnet.apache.org))，\n- PyTorch ([信息](https:\u002F\u002Fpytorch.org))，\n- Chainer ([信息](https:\u002F\u002Fchainer.org))，\n- Keras ([信息](https:\u002F\u002Fkeras.io))，\n- TensorFlow 1.x\u002F2.x ([信息](https:\u002F\u002Fwww.tensorflow.org))。\n\n对于每个支持的框架，都提供了一个仅包含模型的 PIP 包，不包含辅助脚本。包列表如下：\n- [gluoncv2](https:\u002F\u002Fpypi.org\u002Fproject\u002Fgluoncv2) 适用于 Gluon，\n- [pytorchcv](https:\u002F\u002Fpypi.org\u002Fproject\u002Fpytorchcv) 适用于 PyTorch，\n- [chainercv2](https:\u002F\u002Fpypi.org\u002Fproject\u002Fchainercv2) 适用于 Chainer，\n- [kerascv](https:\u002F\u002Fpypi.org\u002Fproject\u002Fkerascv) 适用于 Keras，\n- [tensorflowcv](https:\u002F\u002Fpypi.org\u002Fproject\u002Ftensorflowcv) 适用于 TensorFlow 1.x，\n- [tf2cv](https:\u002F\u002Fpypi.org\u002Fproject\u002Ftf2cv) 适用于 TensorFlow 2.x。\n\n目前，大多数模型首先在 Gluon 上实现，然后再移植到其他框架。部分模型已在以下数据集上进行了预训练：[ImageNet-1K](http:\u002F\u002Fwww.image-net.org)、[CIFAR-10\u002F100](https:\u002F\u002Fwww.cs.toronto.edu\u002F~kriz\u002Fcifar.html)、[SVHN](http:\u002F\u002Fufldl.stanford.edu\u002Fhousenumbers)、[CUB-200-2011](http:\u002F\u002Fwww.vision.caltech.edu\u002Fvisipedia\u002FCUB-200-2011.html)、[Pascal VOC2012](http:\u002F\u002Fhost.robots.ox.ac.uk\u002Fpascal\u002FVOC\u002Fvoc2012)、[ADE20K](http:\u002F\u002Fgroups.csail.mit.edu\u002Fvision\u002Fdatasets\u002FADE20K)、[Cityscapes](https:\u002F\u002Fwww.cityscapes-dataset.com) 和 [COCO](http:\u002F\u002Fcocodataset.org)。所有预训练权重在使用时都会自动加载。有关此类自动加载权重的示例，请参阅各特定包文档中对应的章节：\n- [Gluon 模型](gluon\u002FREADME.md)，\n- [PyTorch 模型](https:\u002F\u002Fgithub.com\u002Fosmr\u002Fpytorchcv\u002Fblob\u002Fmaster\u002FREADME.md)，\n- [Chainer 模型](chainer_\u002FREADME.md)，\n- [Keras 模型](keras_\u002FREADME.md)，\n- [TensorFlow 1.x 模型](tensorflow_\u002FREADME.md)，\n- [TensorFlow 2.x 模型](tensorflow2\u002FREADME.md)。\n\n## 安装\n\n要使用训练\u002F评估脚本以及所有模型，您需要克隆仓库并安装依赖项：\n```\ngit clone git@github.com:osmr\u002Fimgclsmob.git\npip install -r requirements.txt\n```\n\n## 已实现分类模型一览表\n\n几点说明：\n- `Repo` 表示作者的代码仓库（如果存在）。\n- `a`、`b`、`c`、`d` 和 `e` 分别表示该模型在 ImageNet-1K、CIFAR-10、CIFAR-100、SVHN 和 CUB-200-2011 数据集上的实现。\n- `A`、`B`、`C`、`D` 和 `E` 表示对应数据集上存在预训练模型。\n\n| 模型 | [Gluon](gluon\u002FREADME.md) | [PyTorch](https:\u002F\u002Fgithub.com\u002Fosmr\u002Fpytorchcv\u002Fblob\u002Fmaster\u002FREADME.md) | [Chainer](chainer_\u002FREADME.md) | [Keras](keras_\u002FREADME.md) | [TF](tensorflow_\u002FREADME.md) | [TF2](tensorflow2\u002FREADME.md) | 论文 | 仓库 | 年份 |\n| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |\n| AlexNet | A | A | A | A | A | A | [链接](http:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F4824-imagenet-classification-with-deep-convolutional-neural-networks.pdf) | [链接](https:\u002F\u002Fcode.google.com\u002Farchive\u002Fp\u002Fcuda-convnet2) | 2012 |\n| ZFNet | A | A | A | A | A | A | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1311.2901) | - | 2013 |\n| VGG | A | A | A | A | A | A | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1409.1556) | - | 2014 |\n| BN-VGG | A | A | A | A | A | A | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1409.1556) | - | 2015 |\n| BN-Inception | A | A | A | - | - | A | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1502.03167) | - | 2015 |\n| ResNet | ABCDE | ABCDE | ABCDE | A | A | ABCDE | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1512.03385) | [链接](https:\u002F\u002Fgithub.com\u002FKaimingHe\u002Fdeep-residual-networks) | 2015 |\n| PreResNet | ABCD | ABCD | ABCD | A | A | ABCD | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1603.05027) | [链接](https:\u002F\u002Fgithub.com\u002Ffacebook\u002Ffb.resnet.torch) | 2016 |\n| ResNeXt | ABCD | ABCD | ABCD | A | A | ABCD | [链接](http:\u002F\u002Farxiv.org\u002Fabs\u002F1611.05431) | [链接](https:\u002F\u002Fgithub.com\u002Ffacebookresearch\u002FResNeXt) | 2016 |\n| SENet | A | A | A | A | A | A | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1709.01507) | [链接](https:\u002F\u002Fgithub.com\u002Fhujie-frank\u002FSENet) | 2017 |\n| SE-ResNet | ABCDE | ABCDE | ABCDE | A | A | ABCDE | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1709.01507) | [链接](https:\u002F\u002Fgithub.com\u002Fhujie-frank\u002FSENet) | 2017 |\n| SE-PreResNet | ABCD | ABCD | ABCD | A | A | ABCD | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1709.01507) | [链接](https:\u002F\u002Fgithub.com\u002Fhujie-frank\u002FSENet) | 2017 |\n| SE-ResNeXt | A | A | A | A | A | A | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1709.01507) | [链接](https:\u002F\u002Fgithub.com\u002Fhujie-frank\u002FSENet) | 2017 |\n| ResNeSt(A) | A | A | A | - | - | A | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F2004.08955) | [链接](https:\u002F\u002Fgithub.com\u002Fzhanghang1989\u002FResNeSt) | 2020 |\n| IBN-ResNet | A | A | - | - | - | A | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1807.09441) | [链接](https:\u002F\u002Fgithub.com\u002FXingangPan\u002FIBN-Net) | 2018 |\n| IBN-ResNeXt | A | A | - | - | - | A | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1807.09441) | [链接](https:\u002F\u002Fgithub.com\u002FXingangPan\u002FIBN-Net) | 2018 |\n| IBN-DenseNet | A | A | - | - | - | A | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1807.09441) | [链接](https:\u002F\u002Fgithub.com\u002FXingangPan\u002FIBN-Net) | 2018 |\n| AirNet | A | A | A | - | - | A | [链接](https:\u002F\u002Fieeexplore.ieee.org\u002Fdocument\u002F8510896) | [链接](https:\u002F\u002Fgithub.com\u002Fsoeaver\u002FAirNet-PyTorch) | 2018 |\n| AirNeXt | A | A | A | - | - | A | [链接](https:\u002F\u002Fieeexplore.ieee.org\u002Fdocument\u002F8510896) | [链接](https:\u002F\u002Fgithub.com\u002Fsoeaver\u002FAirNet-PyTorch) | 2018 |\n| BAM-ResNet | A | A | A | - | - | A | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1807.06514) | [链接](https:\u002F\u002Fgithub.com\u002FJongchan\u002Fattention-module) | 2018 |\n| CBAM-ResNet | A | A | A | - | - | A | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1807.06521) | [链接](https:\u002F\u002Fgithub.com\u002FJongchan\u002Fattention-module) | 2018 |\n| ResAttNet | a | a | a | - | - | - | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1704.06904) | [链接](https:\u002F\u002Fgithub.com\u002Ffwang91\u002Fresidual-attention-network) | 2017 |\n| SKNet | a | a | a | - | - | - | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1903.06586) | [链接](https:\u002F\u002Fgithub.com\u002Fimplus\u002FSKNet) | 2019 |\n| SCNet | A | A | A | - | - | A | [链接](http:\u002F\u002Fmftp.mmcheng.net\u002FPapers\u002F20cvprSCNet.pdf) | [链接](https:\u002F\u002Fgithub.com\u002FMCG-NKU\u002FSCNet) | 2020 |\n| RegNet | A | A | A | - | - | A | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F2003.13678) | [链接](https:\u002F\u002Fgithub.com\u002Ffacebookresearch\u002Fpycls) | 2020 |\n| DIA-ResNet | aBCD | aBCD | aBCD | - | - | - | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1905.10671) | [链接](https:\u002F\u002Fgithub.com\u002Fgbup-group\u002FDIANet) | 2019 |\n| DIA-PreResNet | aBCD | aBCD | aBCD | - | - | - | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1905.10671) | [链接](https:\u002F\u002Fgithub.com\u002Fgbup-group\u002FDIANet) | 2019 |\n| PyramidNet | ABCD | ABCD | ABCD | - | - | ABCD | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1610.02915) | [链接](https:\u002F\u002Fgithub.com\u002Fjhkim89\u002FPyramidNet) | 2016 |\n| DiracNetV2 | A | A | A | - | - | A | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1706.00388) | [链接](https:\u002F\u002Fgithub.com\u002Fszagoruyko\u002Fdiracnets) | 2017 |\n| ShaResNet | a | a | a | - | - | - | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1702.08782) | [链接](https:\u002F\u002Fgithub.com\u002Faboulch\u002Fsharesnet) | 2017 |\n| CRU-Net | A | - | - | - | - | - | [链接](https:\u002F\u002Fwww.ijcai.org\u002Fproceedings\u002F2018\u002F88) | [链接](https:\u002F\u002Fgithub.com\u002Fcypw\u002FCRU-Net) | 2018 |\n| DenseNet | ABCD | ABCD | ABCD | A | A | ABCD | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1608.06993) | [链接](https:\u002F\u002Fgithub.com\u002Fliuzhuang13\u002FDenseNet) | 2016 |\n| CondenseNet | A | A | A | - | - | - | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1711.09224) | [链接](https:\u002F\u002Fgithub.com\u002FShichenLiu\u002FCondenseNet) | 2017 |\n| SparseNet | a | a | a | - | - | - | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1801.05895) | [链接](https:\u002F\u002Fgithub.com\u002FLyken17\u002FSparseNet) | 2018 |\n| PeleeNet | A | A | A | - | - | A | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1804.06882) | [链接](https:\u002F\u002Fgithub.com\u002FRobert-JunWang\u002FPelee) | 2018 |\n| Oct-ResNet | abcd | a | a | - | - | - | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1904.05049) | - | 2019 |\n| Res2Net | a | - | - | - | - | - | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1904.01169) | - | 2019 |\n| WRN | ABCD | ABCD | ABCD | - | - | a | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1605.07146) | [链接](https:\u002F\u002Fgithub.com\u002Fszagoruyko\u002Fwide-residual-networks) | 2016 |\n| WRN-1bit | BCD | BCD | BCD | - | - | - | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1802.08530) | [链接](https:\u002F\u002Fgithub.com\u002FMcDonnell-Lab\u002F1-bit-per-weight) | 2018 |\n| DRN-C | A | A | A | - | - | A | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1705.09914) | [链接](https:\u002F\u002Fgithub.com\u002Ffyu\u002Fdrn) | 2017 |\n| DRN-D | A | A | A | - | - | A | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1705.09914) | [链接](https:\u002F\u002Fgithub.com\u002Ffyu\u002Fdrn) | 2017 |\n| DPN | A | A | A | - | - | A | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1707.01629) | [链接](https:\u002F\u002Fgithub.com\u002Fcypw\u002FDPNs) | 2017 |\n| DarkNet Ref | A | A | A | A | A | A | [链接](https:\u002F\u002Fgithub.com\u002Fpjreddie\u002Fdarknet) | [链接](https:\u002F\u002Fgithub.com\u002Fpjreddie\u002Fdarknet) | - |\n| DarkNet Tiny | A | A | A | A | A | A | [链接](https:\u002F\u002Fgithub.com\u002Fpjreddie\u002Fdarknet) | [链接](https:\u002F\u002Fgithub.com\u002Fpjreddie\u002Fdarknet) | - |\n| DarkNet-19 | a | a | a | a | a | a | [链接](https:\u002F\u002Fgithub.com\u002Fpjreddie\u002Fdarknet) | [链接](https:\u002F\u002Fgithub.com\u002Fpjreddie\u002Fdarknet) | - |\n| DarkNet-53 | A | A | A | A | A | A | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1804.02767) | [链接](https:\u002F\u002Fgithub.com\u002Fpjreddie\u002Fdarknet) | 2018 |\n| ChannelNet | a | a | a | - | a | - | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1809.01330) | [链接](https:\u002F\u002Fgithub.com\u002FHongyangGao\u002FChannelNets) | 2018 |\n| iSQRT-COV-ResNet | a | a | - | - | - | - | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1712.01034) | [链接](https:\u002F\u002Fgithub.com\u002Fjiangtaoxie\u002Ffast-MPN-COV) | 2017 |\n| RevNet | - | a | - | - | - | - | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1707.04585) | [链接](https:\u002F\u002Fgithub.com\u002Frenmengye\u002Frevnet-public) | 2017 |\n| i-RevNet | A | A | A | - | - | - | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1802.07088) | [链接](https:\u002F\u002Fgithub.com\u002Fjhjacobsen\u002Fpytorch-i-revnet) | 2018 |\n| BagNet | A | A | A | - | - | A | [链接](https:\u002F\u002Fopenreview.net\u002Fpdf?id=SkfMWhAqYQ) | [链接](https:\u002F\u002Fgithub.com\u002Fwielandbrendel\u002Fbag-of-local-features-models) | 2019 |\n| DLA | A | A | A | - | - | A | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1707.06484) | [链接](https:\u002F\u002Fgithub.com\u002Fucbdrive\u002Fdla) | 2017 |\n| MSDNet | a | ab | - | - | - | - | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1703.09844) | [链接](https:\u002F\u002Fgithub.com\u002Fgaohuang\u002FMSDNet) | 2017 |\n| FishNet | A | A | A | - | - | - | [链接](http:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F7356-fishnet-a-versatile-backbone-for-image-region-and-pixel-level-prediction.pdf) | [链接](https:\u002F\u002Fgithub.com\u002Fkevin-ssy\u002FFishNet) | 2018 |\n| ESPNetv2 | A | A | A | - | - | - | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1811.11431) | [链接](https:\u002F\u002Fgithub.com\u002Fsacmehta\u002FESPNetv2) | 2018 |\n| DiCENet | A | A | A | - | - | A | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1906.03516) | [链接](https:\u002F\u002Fgithub.com\u002Fsacmehta\u002FEdgeNets) | 2019 |\n| HRNet | A | A | A | - | - | A | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1908.07919) | [链接](https:\u002F\u002Fgithub.com\u002FHRNet\u002FHRNet-Image-Classification) | 2019 |\n| VoVNet | A | A | A | - | - | A | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1904.09730) | [链接](https:\u002F\u002Fgithub.com\u002Fstigma0617\u002FVoVNet.pytorch) | 2019 |\n| SelecSLS | A | A | A | - | - | A | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1907.00837) | [链接](https:\u002F\u002Fgithub.com\u002Fmehtadushy\u002FSelecSLS-Pytorch) | 2019 |\n| HarDNet | A | A | A | - | - | A | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1909.00948) | [链接](https:\u002F\u002Fgithub.com\u002FPingoLH\u002FPytorch-HarDNet) | 2019 |\n| X-DenseNet | aBCD | aBCD | aBCD | - | - | - | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1711.08757) | [链接](https:\u002F\u002Fgithub.com\u002FDrImpossible\u002FDeep-Expander-Networks) | 2017 |\n| SqueezeNet | A | A | A | A | A | A | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1602.07360) | [链接](https:\u002F\u002Fgithub.com\u002FDeepScale\u002FSqueezeNet) | 2016 |\n| SqueezeResNet | A | A | A | A | A | A | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1602.07360) | - | 2016 |\n| SqueezeNext | A | A | A | A | A | A | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1803.10615) | [链接](https:\u002F\u002Fgithub.com\u002Famirgholami\u002FSqueezeNext) | 2018 |\n| ShuffleNet | A | A | A | A | A | A | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1707.01083) | - | 2017 |\n| ShuffleNetV2 | A | A | A | A | A | A | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1807.11164) | - | 2018 |\n| MENet | A | A | A | A | A | A | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1803.09127) | [链接](https:\u002F\u002Fgithub.com\u002Fclavichord93\u002FMENet) | 2018 |\n| MobileNet | AE | AE | AE | A | A | AE | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1704.04861) | [链接](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodels) | 2017 |\n| FD-MobileNet | A | A | A | A | A | A | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1802.03750) | [链接](https:\u002F\u002Fgithub.com\u002Fclavichord93\u002FFD-MobileNet) | 2018 |\n| MobileNetV2 | A | A | A | A | A | A | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1801.04381) | [链接](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodels) | 2018 |\n| MobileNetV3 | A | A | A | A | - | A | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1905.02244) | [链接](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodels) | 2019 |\n| IGCV3 | A | A | A | A | A | A | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1806.00178) | [链接](https:\u002F\u002Fgithub.com\u002Fhomles11\u002FIGCV3) | 2018 |\n| GhostNet | a | a | a | - | - | a | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1911.11907) | [链接](https:\u002F\u002Fgithub.com\u002Fiamhankai\u002Fghostnet) | 2019 |\n| MnasNet | A | A | A | A | A | A | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1807.11626) | - | 2018 |\n| DARTS | A | A | A | - | - | - | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1806.09055) | [链接](https:\u002F\u002Fgithub.com\u002Fquark0\u002Fdarts) | 2018 |\n| ProxylessNAS | AE | AE | AE | - | - | AE | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1812.00332) | [链接](https:\u002F\u002Fgithub.com\u002Fmit-han-lab\u002FProxylessNAS) | 2018 |\n| FBNet-C | A | A | A | - | - | A | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1812.03443) | - | 2018 |\n| Xception | A | A | A | - | - | A | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1610.02357) | [链接](https:\u002F\u002Fgithub.com\u002Ffchollet\u002Fdeep-learning-models) | 2016 |\n| InceptionV3 | A | A | A | - | - | A | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1512.00567) | [链接](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodels) | 2015 |\n| InceptionV4 | A | A | A | - | - | A | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1602.07261) | [链接](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodels) | 2016 |\n| InceptionResNetV1 | A | A | A | - | - | A | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1602.07261) | [链接](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodels) | 2016 |\n| InceptionResNetV2 | A | A | A | - | - | A | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1602.07261) | [链接](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodels) | 2016 |\n| PolyNet | A | A | A | - | - | A | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1611.05725) | [链接](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fpolynet) | 2016 |\n| NASNet-Large | A | A | A | - | - | A | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1707.07012) | [链接](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodels) | 2017 |\n| NASNet-Mobile | A | A | A | - | - | A | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1707.07012) | [链接](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodels) | 2017 |\n| PNASNet-Large | A | A | A | - | - | A | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1712.00559) | [链接](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodels) | 2017 |\n| SPNASNet | A | A | A | - | - | A | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1904.02877) | [链接](https:\u002F\u002Fgithub.com\u002Fdstamoulis\u002Fsingle-path-nas) | 2019 |\n| EfficientNet | A | A | A | A | - | A | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1905.11946) | [链接](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Ftpu\u002Ftree\u002Fmaster\u002Fmodels\u002Fofficial\u002Fefficientnet) | 2019 |\n| MixNet | A | A | A | - | - | A | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1907.09595) | [链接](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Ftpu\u002Ftree\u002Fmaster\u002Fmodels\u002Fofficial\u002Fmnasnet\u002Fmixnet) | 2019 |\n| NIN | BCD | BCD | BCD | - | - | - | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1312.4400) | [链接](https:\u002F\u002Fgist.github.com\u002Fmavenlin\u002Fe56253735ef32c3c296d) | 2013 |\n| RoR-3 | BCD | BCD | BCD | - | - | - | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1608.02908) | - | 2016 |\n| RiR | BCD | BCD | BCD | - | - | - | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1603.08029) | - | 2016 |\n| ResDrop-ResNet | bcd | bcd | bcd | - | - | - | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1603.09382) | [链接](https:\u002F\u002Fgithub.com\u002Fyueatsprograms\u002FStochastic_Depth) | 2016 |\n| Shake-Shake-ResNet | BCD | BCD | BCD | - | - | - | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1705.07485) | [链接](https:\u002F\u002Fgithub.com\u002Fxgastaldi\u002Fshake-shake) | 2017 |\n| ShakeDrop-ResNet | bcd | bcd | bcd | - | - | - | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1802.02375) | - | 2018 |\n| FractalNet | bc | bc | - | - | - | - | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1605.07648) | [链接](https:\u002F\u002Fgithub.com\u002Fgustavla\u002Ffractalnet) | 2016 |\n| NTS-Net | E | E | E | - | - | - | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1809.00287) | [链接](https:\u002F\u002Fgithub.com\u002Fyangze0930\u002FNTS-Net) | 2018 |\n\n## 已实现的分割模型表格\n\n一些说明：\n- `a\u002FA` 对应 Pascal VOC2012。\n- `b\u002FB` 对应 ADE20K。\n- `c\u002FC` 对应 Cityscapes。\n- `d\u002FD` 对应 COCO。\n- `e\u002FE` 对应 CelebAMask-HQ。\n\n| 模型 | [Gluon](gluon\u002FREADME.md) | [PyTorch](https:\u002F\u002Fgithub.com\u002Fosmr\u002Fpytorchcv\u002Fblob\u002Fmaster\u002FREADME.md) | [Chainer](chainer_\u002FREADME.md) | [Keras](keras_\u002FREADME.md) | [TF](tensorflow_\u002FREADME.md)  | [TF2](tensorflow_\u002FREADME.md) | 论文 | 仓库 | 年份 |\n| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |\n| PSPNet | ABCD | ABCD | ABCD | - | - | ABCD | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1612.01105) | - | 2016 |\n| DeepLabv3 | ABcD | ABcD | ABcD | - | - | ABcD | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1706.05587) | - | 2017 |\n| FCN-8s(d) | ABcD | ABcD | ABcD | - | - | ABcD | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1411.4038) | - | 2014 |\n| ICNet | C | C | C | - | - | C | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1704.08545) | [链接](https:\u002F\u002Fgithub.com\u002Fhszhao\u002FICNet) | 2017 |\n| SINet | C | C | C | - | - | c | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1911.09099) | [链接](https:\u002F\u002Fgithub.com\u002Fclovaai\u002Fc3_sinet) | 2019 |\n| BiSeNet | e | e | e | - | - | e | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1808.00897) | - | 2018 |\n| DANet | C | C | C | - | - | C | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1809.02983) | [链接](https:\u002F\u002Fgithub.com\u002Fjunfu1115\u002FDANet) | 2018 |\n| Fast-SCNN | C | C | C | - | - | C | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1902.04502) | - | 2019 |\n| CGNet | c | c | c | - | - | c | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1811.08201) | [链接](https:\u002F\u002Fgithub.com\u002FwutianyiRosun\u002FCGNet) | 2018 |\n| DABNet | c | c | c | - | - | c | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1907.11357) | [链接](https:\u002F\u002Fgithub.com\u002FReagan1311\u002FDABNet) | 2019 |\n| FPENet | c | c | c | - | - | c | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1909.08599) | - | 2019 |\n| ContextNet | - | c | - | - | - | - | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1805.04554) | - | 2018 |\n| LEDNet | c | c | c | - | - | c | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1905.02423) | - | 2019 |\n| ESNet | - | c | - | - | - | - | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1906.09826) | - | 2019 |\n| EDANet | - | c | - | - | - | - | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1809.06323) | [链接](https:\u002F\u002Fgithub.com\u002Fshaoyuanlo\u002FEDANet) | 2018 |\n| ENet | - | c | - | - | - | - | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1606.02147) | - | 2016 |\n| ERFNet | - | c | - | - | - | - | [链接](http:\u002F\u002Fwww.robesafe.uah.es\u002Fpersonal\u002Feduardo.romera\u002Fpdfs\u002FRomera17tits.pdf) | - | 2017 |\n| LinkNet | - | c | - | - | - | - | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1707.03718) | - | 2017 |\n| SegNet | - | c | - | - | - | - | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1511.00561) | - | 2015 |\n| U-Net | - | c | - | - | - | - | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1505.04597) | - | 2015 |\n| SQNet | - | c | - | - | - | - | [链接](https:\u002F\u002Fopenreview.net\u002Fpdf?id=S1uHiFyyg) | - | 2016 |\n\n## 已实现的对象检测模型表格\n\n一些说明：\n- `a\u002FA` 对应 COCO。\n\n| 模型 | [Gluon](gluon\u002FREADME.md) | [PyTorch](https:\u002F\u002Fgithub.com\u002Fosmr\u002Fpytorchcv\u002Fblob\u002Fmaster\u002FREADME.md) | [Chainer](chainer_\u002FREADME.md) | [Keras](keras_\u002FREADME.md) | [TF](tensorflow_\u002FREADME.md)  | [TF2](tensorflow2\u002FREADME.md) | 论文 | 仓库 | 年份 |\n| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |\n| CenterNet | a | a | a | - | - | a | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1904.07850) | [链接](https:\u002F\u002Fgithub.com\u002Fxingyizhou\u002FCenterNet) | 2019 |\n\n## 已实现的人体姿态估计模型表格\n\n一些说明：\n- `a\u002FA` 对应 COCO。\n\n| 模型 | [Gluon](gluon\u002FREADME.md) | [PyTorch](https:\u002F\u002Fgithub.com\u002Fosmr\u002Fpytorchcv\u002Fblob\u002Fmaster\u002FREADME.md) | [Chainer](chainer_\u002FREADME.md) | [Keras](keras_\u002FREADME.md) | [TF](tensorflow_\u002FREADME.md)  | [TF2](tensorflow2\u002FREADME.md) | 论文 | 仓库 | 年份 |\n| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |\n| AlphaPose | A | A | A | - | - | A | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1612.00137) | [链接](https:\u002F\u002Fgithub.com\u002FMVIG-SJTU\u002FAlphaPose) | 2016 |\n| SimplePose | A | A | A | - | - | A | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1804.06208) | [链接](https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Fhuman-pose-estimation.pytorch) | 2018 |\n| SimplePose(Mobile) | A | A | A | - | - | A | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1804.06208) | - | 2018 |\n| Lightweight OpenPose | A | A | A | - | - | A | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1811.12004) | [链接](https:\u002F\u002Fgithub.com\u002FDaniil-Osokin\u002Flightweight-human-pose-estimation-3d-demo.pytorch) | 2018 |\n| IBPPose | A | A | A | - | - | A | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1911.10529) | [链接](https:\u002F\u002Fgithub.com\u002Fjialee93\u002FImproved-Body-Parts) | 2019 |\n\n## 已实现的自动语音识别模型表格\n\n一些说明：\n- `a\u002FA` 对应 LibriSpeech。\n- `b\u002FB` 对应 Mozilla Common Voice。\n\n| 模型 | [Gluon](gluon\u002FREADME.md) | [PyTorch](https:\u002F\u002Fgithub.com\u002Fosmr\u002Fpytorchcv\u002Fblob\u002Fmaster\u002FREADME.md) | [Chainer](chainer_\u002FREADME.md) | [Keras](keras_\u002FREADME.md) | [TF](tensorflow_\u002FREADME.md)  | [TF2](tensorflow2\u002FREADME.md) | 论文 | 仓库 | 年份 |\n| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |\n| Jasper DR | AB | AB | ab | - | - | ab | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1904.03288) | [链接](https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FNeMo) | 2019 |\n| QuartzNet | AB | AB | ab | - | - | ab | [链接](https:\u002F\u002Farxiv.org\u002Fabs\u002F1910.10261) | [链接](https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FNeMo) | 2019 |","# imgclsmob 快速上手指南\n\n`imgclsmob` 是一个专注于计算机视觉任务的深度学习网络研究仓库，提供了多种分类、分割、检测和姿态估计模型的实现。该项目的特点是**同一套模型架构支持多种主流框架**（MXNet\u002FGluon, PyTorch, Chainer, Keras, TensorFlow 1.x\u002F2.x），并自动加载在 ImageNet、CIFAR 等数据集上的预训练权重。\n\n## 环境准备\n\n在开始之前，请确保您的开发环境满足以下要求：\n\n*   **操作系统**：Linux, macOS 或 Windows\n*   **Python 版本**：推荐 Python 3.10（兼容 3.6+）\n*   **深度学习框架**：根据您选择的框架安装对应版本（如 `torch`, `tensorflow`, `mxnet` 等）。\n    *   *建议*：国内用户可使用清华源或阿里源加速框架安装。例如安装 PyTorch：\n        ```bash\n        pip install torch torchvision -i https:\u002F\u002Fpypi.tuna.tsinghua.edu.cn\u002Fsimple\n        ```\n*   **Git**：用于克隆代码仓库。\n\n## 安装步骤\n\n本项目提供两种使用方式：直接安装对应的 PIP 包（仅含模型）或克隆完整仓库（含训练\u002F评估脚本）。若要体验完整功能，推荐克隆仓库。\n\n### 1. 克隆仓库\n```bash\ngit clone git@github.com:osmr\u002Fimgclsmob.git\ncd imgclsmob\n```\n*注：若 GitHub 连接缓慢，可使用国内镜像（如 Gitee 镜像，如有）或通过代理加速。*\n\n### 2. 安装依赖\n使用 `pip` 安装项目所需的通用依赖库。国内用户建议指定清华源以提升下载速度：\n\n```bash\npip install -r requirements.txt -i https:\u002F\u002Fpypi.tuna.tsinghua.edu.cn\u002Fsimple\n```\n\n### 3. (可选) 安装特定框架的独立包\n如果您只需要在代码中调用模型而不需要训练脚本，可以直接安装针对特定框架优化的 PIP 包：\n\n*   **PyTorch**: `pip install pytorchcv`\n*   **TensorFlow 2.x**: `pip install tf2cv`\n*   **MXNet\u002FGluon**: `pip install gluoncv2`\n*   **Keras**: `pip install kerascv`\n*   **Chainer**: `pip install chainercv2`\n*   **TensorFlow 1.x**: `pip install tensorflowcv`\n\n## 基本使用\n\n`imgclsmob` 的核心优势在于能够自动下载并加载预训练权重。以下以 **PyTorch** 框架为例，演示如何加载一个预训练的 ResNet-50 模型并进行简单的推理。其他框架用法类似。\n\n### 示例：加载预训练模型并推理\n\n```python\nimport torch\nfrom pytorchcv.model_provider import get_model as ptcv_get_model\n\n# 1. 获取模型\n# 参数 'resnet50_v1b' 为模型名称，pretrained=True 表示自动下载并加载 ImageNet 预训练权重\nnet = ptcv_get_model(\"resnet50_v1b\", pretrained=True)\n\n# 2. 设置模型为评估模式\nnet.eval()\n\n# 3. 准备输入数据 (模拟一张 224x224 的图片)\n# 实际使用时请替换为您的图像预处理逻辑\ninput_data = torch.randn(1, 3, 224, 224)\n\n# 4. 执行前向传播\nwith torch.no_grad():\n    output = net(input_data)\n\n# 5. 获取预测结果 (类别概率)\nprobabilities = torch.softmax(output, dim=1)\nprint(f\"输出形状：{output.shape}\")\nprint(f\"最大概率类别索引：{torch.argmax(probabilities, dim=1).item()}\")\n```\n\n### 支持的模型\n该项目支持大量经典与现代模型，包括但不限于：\n*   **基础系列**: AlexNet, VGG, ResNet, DenseNet\n*   **注意力机制**: SENet, CBAM, BAM\n*   **高效网络**: MobileNet (需确认具体变体), ShuffleNet, SqueezeNet, HarDNet\n*   **最新架构**: ResNeSt, RegNet, HRNet, ViT (部分支持)\n\n您可以在各框架对应的文档目录（如 `pytorchcv\u002FREADME.md`）中查询完整的模型列表及具体名称字符串。","某计算机视觉初创团队正致力于开发一款跨框架的工业缺陷检测系统，需要在 PyTorch 和 TensorFlow 2.x 之间快速验证不同骨干网络的性能。\n\n### 没有 imgclsmob 时\n- **重复造轮子耗时**：团队成员需分别为 PyTorch 和 TensorFlow 手动复现 ResNet、MobileNet 等经典模型，代码维护成本极高且容易引入实现误差。\n- **预训练权重难寻**：难以找到针对特定数据集（如 CUB-200 或 SVHN）的统一预训练权重，导致模型收敛慢，实验周期被大幅拉长。\n- **框架迁移困难**：当需要将验证好的 Gluon 模型迁移至生产环境常用的 TensorFlow 2.x 时，缺乏标准化的转换脚本，重构过程充满不确定性。\n- **基准对比混乱**：由于各框架下的模型实现细节不一致，无法公平对比不同架构在相同数据集上的真实性能，阻碍了技术选型决策。\n\n### 使用 imgclsmob 后\n- **一键调用多框架模型**：直接通过 `pytorchcv` 或 `tf2cv` pip 包 instant 加载几十种已实现的分类与分割模型，无需编写底层网络结构代码。\n- **自动加载精准权重**：系统自动下载并加载已在 ImageNet-1K、CIFAR 等权威数据集上训练好的权重，显著加速了下游缺陷检测任务的微调过程。\n- **无缝跨框架移植**：利用其统一的模型定义，团队可轻松将原型从研究用的 Gluon\u002FPyTorch 切换至部署友好的 TensorFlow 2.x，确保逻辑一致。\n- **标准化性能评估**：基于同一套经过严格验证的实现代码，团队能在不同框架间进行公平的精度与速度基准测试，快速锁定最优架构。\n\nimgclsmob 通过提供统一、预训练且多框架支持的深度学习模型库，将算法团队的研发重心从繁琐的代码复现转移到了核心业务逻辑的创新上。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fosmr_imgclsmob_552a0f6f.png","osmr","Oleg Sémery","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002Fosmr_2489b34b.jpg","Mathematician",null,"osemery@gmail.com","https:\u002F\u002Fgithub.com\u002Fosmr",[81,85,89],{"name":82,"color":83,"percentage":84},"Python","#3572A5",100,{"name":86,"color":87,"percentage":88},"Shell","#89e051",0,{"name":90,"color":91,"percentage":88},"Dockerfile","#384d54",3016,557,"2026-04-02T08:37:32","MIT","未说明","未说明（取决于所选框架及具体模型，训练通常需要 GPU，推理可选 CPU）",{"notes":99,"python":100,"dependencies":101},"该仓库支持多种深度学习框架（MXNet, PyTorch, Chainer, Keras, TensorFlow），用户需根据选择的框架安装对应依赖。预训练权重在使用时会自动下载。若仅需使用模型而非训练脚本，可安装对应的独立 PIP 包（如 pytorchcv, gluoncv2 等）。具体硬件需求取决于所选模型的大小和任务类型。","3.10",[102,103,104,105,106],"MXNet\u002FGluon","PyTorch","Chainer","Keras","TensorFlow 1.x\u002F2.x",[14,108,15],"其他",[110,111,112,113,114,115,116,117,118,119,120,121,122,123,124,125,126,127,128],"machine-learning","deep-learning","mxnet","gluon","pytorch","classification","imagenet","neural-network","image-classification","chainer","keras","tensorflow","pretrained-models","cifar","segmentation","tensorflow2","semantic-segmentation","human-pose-estimation","3d-face-reconstruction","2026-03-27T02:49:30.150509","2026-04-07T00:50:23.751016",[132,137,142,147,152,157],{"id":133,"question_zh":134,"answer_zh":135,"source_url":136},20520,"使用 PSPNet 等分割模型训练时，遇到'Expected more than 1 value per channel when training'错误怎么办？","该错误通常是因为批量大小（batch_size）设置为 1，导致批归一化层无法计算统计值。解决方法是将 batch_size 设置为 2 或更大。例如：`sample_batch_size = 2`。","https:\u002F\u002Fgithub.com\u002Fosmr\u002Fimgclsmob\u002Fissues\u002F41",{"id":138,"question_zh":139,"answer_zh":140,"source_url":141},20521,"SimplePose Mobile 模型在 TensorFlow 中训练时报错，而普通 SimplePose 模型正常，如何解决？","这通常与 Pixel Shuffle 层的实现有关。可以尝试应用特定的修复提交（commit e0ce2cf3ed3b6a9702d5c46568aa48258196ed53）。或者，手动实现兼容的 Pixel Shuffle 层，使用 `tf.nn.depth_to_space` 函数：\n```python\nclass PixelShuffle(nn.Layer):\n    def __init__(self, scale_factor, data_format=\"channels_last\", **kwargs):\n        super(PixelShuffle, self).__init__(**kwargs)\n        self.scale_factor = scale_factor\n        self.data_format = data_format\n\n    def call(self, x, training=None):\n        if is_channels_first(self.data_format):\n            x = tf.nn.depth_to_space(input=x, block_size=self.scale_factor, data_format='NCHW')          \n        else:\n            x = tf.nn.depth_to_space(input=x, block_size=self.scale_factor, data_format='NHWC')\n        return x\n```","https:\u002F\u002Fgithub.com\u002Fosmr\u002Fimgclsmob\u002Fissues\u002F55",{"id":143,"question_zh":144,"answer_zh":145,"source_url":146},20522,"使用预训练的 PSPNet 模型（如 Cityscapes 版本）时，输出是一个包含两个张量的元组，它们分别代表什么？如何只获取一个输出以节省计算时间？","输出为元组通常是因为模型配置中包含用于训练辅助损失的额外头（auxiliary head）。如果在推理时只需要主输出，请确保在加载模型时不要启用仅用于训练的参数（如 `pretrained_backbone=True` 通常用于训练场景）。维护者指出 `pretrained_backbone` 选项主要用于训练过程。若需单一输出，请检查模型定义或仅提取元组中的第一个元素（主预测结果）。","https:\u002F\u002Fgithub.com\u002Fosmr\u002Fimgclsmob\u002Fissues\u002F92",{"id":148,"question_zh":149,"answer_zh":150,"source_url":151},20523,"ResNeSt 模型在 TensorFlow 中出现'TypeError: Failed to convert object of type tuple to Tensor'错误，如何解决？","这是模型代码中的一个已知问题。请更新仓库到最新版本的代码，测试最后一次提交（last commit）后的效果，该提交已包含对此问题的修复补丁。","https:\u002F\u002Fgithub.com\u002Fosmr\u002Fimgclsmob\u002Fissues\u002F86",{"id":153,"question_zh":154,"answer_zh":155,"source_url":156},20524,"在使用 SqueezeNext 模型配合 Keras\u002FTensorFlow 进行训练并保存回调（ModelCheckpoint）时，出现'TypeError: Not JSON Serializable'错误，如何处理？","该错误通常由环境包版本不兼容引起。建议更新所有相关的依赖包，包括 keras、tensorflow 和 kerascv 到最新版本，以确保模型序列化和回调功能的兼容性。","https:\u002F\u002Fgithub.com\u002Fosmr\u002Fimgclsmob\u002Fissues\u002F38",{"id":158,"question_zh":159,"answer_zh":160,"source_url":161},20525,"为什么找不到 DPN（Dual Path Network）模型的预训练权重？","在该仓库的当前版本中，可能尚未提供 DPN 系列模型（如 DPN98, DPN68）的预训练权重。如果在使用 fastai 等框架调用时报错提示找不到权重，说明这些特定架构的预训练文件暂未发布或不支持直接通过该接口加载。","https:\u002F\u002Fgithub.com\u002Fosmr\u002Fimgclsmob\u002Fissues\u002F56",[163,168,173,178,183,188,193,198,203,208,213,218,223,228,233,238,243,248,253,258],{"id":164,"version":165,"summary_zh":166,"released_at":167},126508,"v0.0.570","用于光流估计的 RAFT 小型\u002F轻量级模型。  \n用于视频修复的 ProPainter 和 ProPainter-RFC 模型。","2024-08-14T20:39:47",{"id":169,"version":170,"summary_zh":171,"released_at":172},126509,"v0.0.569","在额外训练后，ImageNet-1K 数据集上的 ResNet-10、DiCENet x2.0 和 ESPNetv2 x1.0 模型","2021-12-14T10:22:55",{"id":174,"version":175,"summary_zh":176,"released_at":177},126510,"v0.0.568","在额外训练后用于 ImageNet-1K 数据集的 ESPNetv2 x1.0 模型","2021-10-29T06:55:17",{"id":179,"version":180,"summary_zh":181,"released_at":182},126511,"v0.0.567","在额外训练后，DiCENet x0.75 和 ESPNetv2 x0.5 模型在 ImageNet-1K 数据集上的表现","2021-10-26T06:11:48",{"id":184,"version":185,"summary_zh":186,"released_at":187},126512,"v0.0.566","在额外训练后，ESPNetv2 x1.5 和 x2.0 模型在 ImageNet-1K 数据集上的表现","2021-10-01T08:59:50",{"id":189,"version":190,"summary_zh":191,"released_at":192},126513,"v0.0.565","在额外训练后用于 ImageNet-1K 数据集的 ESPNetv2 x2.0 模型","2021-09-28T07:51:21",{"id":194,"version":195,"summary_zh":196,"released_at":197},126514,"v0.0.564","在额外训练后，用于 ImageNet-1K 的几个预训练模型：\nDRN-D-105、ESPNetv2 x1.25 和 i-RevNet-301。","2021-09-28T00:27:20",{"id":199,"version":200,"summary_zh":201,"released_at":202},126515,"v0.0.563","在额外训练后，DiCENet x0.2 模型在 ImageNet-1K 上的表现","2021-09-23T19:49:42",{"id":204,"version":205,"summary_zh":206,"released_at":207},126516,"v0.0.562","在额外训练后用于 ImageNet-1K 数据集的 ESPNetv2 x0.5 模型","2021-09-21T08:40:45",{"id":209,"version":210,"summary_zh":211,"released_at":212},126517,"v0.0.561","在额外训练后，针对 ImageNet-1K 的几个预训练模型包括：BagNet-33、DiCENet x0.5 以及 SE-ResNeXt-101 (64x4d)。","2021-09-20T05:52:12",{"id":214,"version":215,"summary_zh":216,"released_at":217},126518,"v0.0.560","ESPNetv2 x0.5 models for ImageNet-1K after extra training","2021-09-03T08:31:05",{"id":219,"version":220,"summary_zh":221,"released_at":222},126519,"v0.0.559","DRN-C-58 models for ImageNet-1K after extra training","2021-09-01T08:05:24",{"id":224,"version":225,"summary_zh":226,"released_at":227},126520,"v0.0.558","BagNet-17 & CRU-Net-56 models for ImageNet-1K after extra training","2021-08-31T07:40:50",{"id":229,"version":230,"summary_zh":231,"released_at":232},126521,"v0.0.557","HarDNet-68 models for ImageNet-1K after extra training","2021-08-27T06:33:05",{"id":234,"version":235,"summary_zh":236,"released_at":237},126522,"v0.0.556","DRN-C-42 models for ImageNet-1K after extra training","2021-08-25T18:30:52",{"id":239,"version":240,"summary_zh":241,"released_at":242},126523,"v0.0.555","Several QuartzNet & Jasper DR pretrained models for Mozilla Common Voice reconverted with integrated preprocessing weights","2021-08-25T10:44:47",{"id":244,"version":245,"summary_zh":246,"released_at":247},126524,"v0.0.554","DRN-D-54 models for ImageNet-1K after extra training","2021-08-23T08:37:28",{"id":249,"version":250,"summary_zh":251,"released_at":252},126525,"v0.0.553","BagNet-9 & IBN-ResNeXt-101 (32x4d) models for ImageNet-1K after extra training","2021-08-22T10:04:41",{"id":254,"version":255,"summary_zh":256,"released_at":257},126526,"v0.0.552","Several pretrained models for ImageNet-1K after extra training:\r\nInceptionV3, InceptionResNetV1, IBN-ResNet-101, IBN(b)-ResNet-50, DRN-D-38.","2021-08-19T00:13:05",{"id":259,"version":260,"summary_zh":261,"released_at":262},126527,"v0.0.551","VoVNet-27-slim & InceptionV3 models for ImageNet-1K are ready to use","2021-07-26T11:01:11"]