[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"similar-CMU-Perceptual-Computing-Lab--openpose":3,"tool-CMU-Perceptual-Computing-Lab--openpose":64},[4,17,27,35,43,56],{"id":5,"name":6,"github_repo":7,"description_zh":8,"stars":9,"difficulty_score":10,"last_commit_at":11,"category_tags":12,"status":16},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。",162132,3,"2026-04-05T11:01:52",[13,14,15],"开发框架","图像","Agent","ready",{"id":18,"name":19,"github_repo":20,"description_zh":21,"stars":22,"difficulty_score":23,"last_commit_at":24,"category_tags":25,"status":16},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",138956,2,"2026-04-05T11:33:21",[13,15,26],"语言模型",{"id":28,"name":29,"github_repo":30,"description_zh":31,"stars":32,"difficulty_score":23,"last_commit_at":33,"category_tags":34,"status":16},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。",107662,"2026-04-03T11:11:01",[13,14,15],{"id":36,"name":37,"github_repo":38,"description_zh":39,"stars":40,"difficulty_score":23,"last_commit_at":41,"category_tags":42,"status":16},3704,"NextChat","ChatGPTNextWeb\u002FNextChat","NextChat 是一款轻量且极速的 AI 助手，旨在为用户提供流畅、跨平台的大模型交互体验。它完美解决了用户在多设备间切换时难以保持对话连续性，以及面对众多 AI 模型不知如何统一管理的痛点。无论是日常办公、学习辅助还是创意激发，NextChat 都能让用户随时随地通过网页、iOS、Android、Windows、MacOS 或 Linux 端无缝接入智能服务。\n\n这款工具非常适合普通用户、学生、职场人士以及需要私有化部署的企业团队使用。对于开发者而言，它也提供了便捷的自托管方案，支持一键部署到 Vercel 或 Zeabur 等平台。\n\nNextChat 的核心亮点在于其广泛的模型兼容性，原生支持 Claude、DeepSeek、GPT-4 及 Gemini Pro 等主流大模型，让用户在一个界面即可自由切换不同 AI 能力。此外，它还率先支持 MCP（Model Context Protocol）协议，增强了上下文处理能力。针对企业用户，NextChat 提供专业版解决方案，具备品牌定制、细粒度权限控制、内部知识库整合及安全审计等功能，满足公司对数据隐私和个性化管理的高标准要求。",87618,"2026-04-05T07:20:52",[13,26],{"id":44,"name":45,"github_repo":46,"description_zh":47,"stars":48,"difficulty_score":23,"last_commit_at":49,"category_tags":50,"status":16},2268,"ML-For-Beginners","microsoft\u002FML-For-Beginners","ML-For-Beginners 是由微软推出的一套系统化机器学习入门课程，旨在帮助零基础用户轻松掌握经典机器学习知识。这套课程将学习路径规划为 12 周，包含 26 节精炼课程和 52 道配套测验，内容涵盖从基础概念到实际应用的完整流程，有效解决了初学者面对庞大知识体系时无从下手、缺乏结构化指导的痛点。\n\n无论是希望转型的开发者、需要补充算法背景的研究人员，还是对人工智能充满好奇的普通爱好者，都能从中受益。课程不仅提供了清晰的理论讲解，还强调动手实践，让用户在循序渐进中建立扎实的技能基础。其独特的亮点在于强大的多语言支持，通过自动化机制提供了包括简体中文在内的 50 多种语言版本，极大地降低了全球不同背景用户的学习门槛。此外，项目采用开源协作模式，社区活跃且内容持续更新，确保学习者能获取前沿且准确的技术资讯。如果你正寻找一条清晰、友好且专业的机器学习入门之路，ML-For-Beginners 将是理想的起点。",84991,"2026-04-05T10:45:23",[14,51,52,53,15,54,26,13,55],"数据工具","视频","插件","其他","音频",{"id":57,"name":58,"github_repo":59,"description_zh":60,"stars":61,"difficulty_score":10,"last_commit_at":62,"category_tags":63,"status":16},3128,"ragflow","infiniflow\u002Fragflow","RAGFlow 是一款领先的开源检索增强生成（RAG）引擎，旨在为大语言模型构建更精准、可靠的上下文层。它巧妙地将前沿的 RAG 技术与智能体（Agent）能力相结合，不仅支持从各类文档中高效提取知识，还能让模型基于这些知识进行逻辑推理和任务执行。\n\n在大模型应用中，幻觉问题和知识滞后是常见痛点。RAGFlow 通过深度解析复杂文档结构（如表格、图表及混合排版），显著提升了信息检索的准确度，从而有效减少模型“胡编乱造”的现象，确保回答既有据可依又具备时效性。其内置的智能体机制更进一步，使系统不仅能回答问题，还能自主规划步骤解决复杂问题。\n\n这款工具特别适合开发者、企业技术团队以及 AI 研究人员使用。无论是希望快速搭建私有知识库问答系统，还是致力于探索大模型在垂直领域落地的创新者，都能从中受益。RAGFlow 提供了可视化的工作流编排界面和灵活的 API 接口，既降低了非算法背景用户的上手门槛，也满足了专业开发者对系统深度定制的需求。作为基于 Apache 2.0 协议开源的项目，它正成为连接通用大模型与行业专有知识之间的重要桥梁。",77062,"2026-04-04T04:44:48",[15,14,13,26,54],{"id":65,"github_repo":66,"name":67,"description_en":68,"description_zh":69,"ai_summary_zh":69,"readme_en":70,"readme_zh":71,"quickstart_zh":72,"use_case_zh":73,"hero_image_url":74,"owner_login":75,"owner_name":75,"owner_avatar_url":76,"owner_bio":77,"owner_company":77,"owner_location":77,"owner_email":77,"owner_twitter":77,"owner_website":77,"owner_url":78,"languages":79,"stars":107,"forks":108,"last_commit_at":109,"license":110,"difficulty_score":111,"env_os":112,"env_gpu":113,"env_ram":114,"env_deps":115,"category_tags":128,"github_topics":129,"view_count":23,"oss_zip_url":77,"oss_zip_packed_at":77,"status":16,"created_at":149,"updated_at":150,"faqs":151,"releases":180},3760,"CMU-Perceptual-Computing-Lab\u002Fopenpose","openpose","OpenPose: Real-time multi-person keypoint detection library for body, face, hands, and foot estimation","OpenPose 是一款强大的开源实时多人姿态估计库，能够精准识别人体的躯干、面部、手部及脚部关键点位。它成功解决了在单张图像或视频流中，同时对多人进行全身（共计 135 个关键点）高精度检测的技术难题，即使在人物重叠或动作复杂的场景下也能保持稳定的追踪效果。\n\n作为全球首个实现实时全身多人体态捕捉的系统，OpenPose 不仅支持 2D 姿态估算，还具备 3D 姿态重建能力，并提供了便捷的 Unity 插件以赋能游戏开发与虚拟现实应用。其核心优势在于将身体、手脚和面部的检测整合于统一框架内，实现了速度与精度的卓越平衡。\n\n这款工具非常适合计算机视觉研究人员、AI 开发者、动画设计师以及需要人机交互解决方案的工程师使用。无论是用于学术算法研究、影视动作捕捉、体感游戏开发，还是构建智能监控与行为分析系统，OpenPose 都能提供坚实可靠的技术底座，帮助用户轻松将前沿的姿态识别技术落地到实际项目中。","\u003Cdiv align=\"center\">\n    \u003Cimg src=\".github\u002FLogo_main_black.png\" width=\"300\">\n\u003C\u002Fdiv>\n\n-----------------\n\n| **Build Type**   |`Linux`           |`MacOS`           |`Windows`         |\n| :---:            | :---:            | :---:            | :---:            |\n| **Build Status** | [![Status](https:\u002F\u002Fgithub.com\u002FCMU-Perceptual-Computing-Lab\u002Fopenpose\u002Fworkflows\u002FCI\u002Fbadge.svg)](https:\u002F\u002Fgithub.com\u002FCMU-Perceptual-Computing-Lab\u002Fopenpose\u002Factions) | [![Status](https:\u002F\u002Fgithub.com\u002FCMU-Perceptual-Computing-Lab\u002Fopenpose\u002Fworkflows\u002FCI\u002Fbadge.svg)](https:\u002F\u002Fgithub.com\u002FCMU-Perceptual-Computing-Lab\u002Fopenpose\u002Factions) | [![Status](https:\u002F\u002Fci.appveyor.com\u002Fapi\u002Fprojects\u002Fstatus\u002F5leescxxdwen77kg\u002Fbranch\u002Fmaster?svg=true)](https:\u002F\u002Fci.appveyor.com\u002Fproject\u002Fgineshidalgo99\u002Fopenpose\u002Fbranch\u002Fmaster) |\n\n[**OpenPose**](https:\u002F\u002Fgithub.com\u002FCMU-Perceptual-Computing-Lab\u002Fopenpose) has represented the **first real-time multi-person system to jointly detect human body, hand, facial, and foot keypoints (in total 135 keypoints) on single images**.\n\nIt is **authored by** [**Ginés Hidalgo**](https:\u002F\u002Fwww.gineshidalgo.com), [**Zhe Cao**](https:\u002F\u002Fpeople.eecs.berkeley.edu\u002F~zhecao), [**Tomas Simon**](http:\u002F\u002Fwww.cs.cmu.edu\u002F~tsimon), [**Shih-En Wei**](https:\u002F\u002Fscholar.google.com\u002Fcitations?user=sFQD3k4AAAAJ&hl=en), [**Yaadhav Raaj**](https:\u002F\u002Fwww.raaj.tech), [**Hanbyul Joo**](https:\u002F\u002Fjhugestar.github.io), **and** [**Yaser Sheikh**](http:\u002F\u002Fwww.cs.cmu.edu\u002F~yaser). It is **maintained by** [**Ginés Hidalgo**](https:\u002F\u002Fwww.gineshidalgo.com) **and** [**Yaadhav Raaj**](https:\u002F\u002Fwww.raaj.tech). OpenPose would not be possible without the [**CMU Panoptic Studio dataset**](http:\u002F\u002Fdomedb.perception.cs.cmu.edu). We would also like to thank all the people who [have helped OpenPose in any way](doc\u002F09_authors_and_contributors.md).\n\n\n\n\u003Cp align=\"center\">\n    \u003Cimg src=\".github\u002Fmedia\u002Fpose_face_hands.gif\" width=\"480\">\n    \u003Cbr>\n    \u003Csup>Authors \u003Ca href=\"https:\u002F\u002Fwww.gineshidalgo.com\" target=\"_blank\">Ginés Hidalgo\u003C\u002Fa> (left) and \u003Ca href=\"https:\u002F\u002Fjhugestar.github.io\" target=\"_blank\">Hanbyul Joo\u003C\u002Fa> (right) in front of the \u003Ca href=\"http:\u002F\u002Fdomedb.perception.cs.cmu.edu\" target=\"_blank\">CMU Panoptic Studio\u003C\u002Fa>\u003C\u002Fsup>\n\u003C\u002Fp>\n\n\n\n## Contents\n1. [Results](#results)\n2. [Features](#features)\n3. [Related Work](#related-work)\n4. [Installation](#installation)\n5. [Quick Start Overview](#quick-start-overview)\n6. [Send Us Feedback!](#send-us-feedback)\n7. [Citation](#citation)\n8. [License](#license)\n\n\n\n## Results\n### Whole-body (Body, Foot, Face, and Hands) 2D Pose Estimation\n\u003Cp align=\"center\">\n    \u003Cimg src=\".github\u002Fmedia\u002Fdance_foot.gif\" width=\"300\">\n    \u003Cimg src=\".github\u002Fmedia\u002Fpose_face.gif\" width=\"300\">\n    \u003Cimg src=\".github\u002Fmedia\u002Fpose_hands.gif\" width=\"300\">\n    \u003Cbr>\n    \u003Csup>Testing OpenPose: (Left) \u003Ca href=\"https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=2DiQUX11YaY\" target=\"_blank\">\u003Ci>Crazy Uptown Funk flashmob in Sydney\u003C\u002Fi>\u003C\u002Fa> video sequence. (Center and right) Authors \u003Ca href=\"https:\u002F\u002Fwww.gineshidalgo.com\" target=\"_blank\">Ginés Hidalgo\u003C\u002Fa> and \u003Ca href=\"http:\u002F\u002Fwww.cs.cmu.edu\u002F~tsimon\" target=\"_blank\">Tomas Simon\u003C\u002Fa> testing face and hands\u003C\u002Fsup>\n\u003C\u002Fp>\n\n### Whole-body 3D Pose Reconstruction and Estimation\n\u003Cp align=\"center\">\n    \u003Cimg src=\".github\u002Fmedia\u002Fopenpose3d.gif\" width=\"360\">\n    \u003Cbr>\n    \u003Csup>\u003Ca href=\"https:\u002F\u002Fziutinyat.github.io\u002F\" target=\"_blank\">Tianyi Zhao\u003C\u002Fa> testing the OpenPose 3D Module\u003C\u002Fa>\u003C\u002Fsup>\n\u003C\u002Fp>\n\n### Unity Plugin\n\u003Cp align=\"center\">\n    \u003Cimg src=\".github\u002Fmedia\u002Funity_main.png\" width=\"300\">\n    \u003Cimg src=\".github\u002Fmedia\u002Funity_body_foot.png\" width=\"300\">\n    \u003Cimg src=\".github\u002Fmedia\u002Funity_hand_face.png\" width=\"300\">\n    \u003Cbr>\n    \u003Csup>\u003Ca href=\"https:\u002F\u002Fziutinyat.github.io\u002F\" target=\"_blank\">Tianyi Zhao\u003C\u002Fa> and \u003Ca href=\"https:\u002F\u002Fwww.gineshidalgo.com\" target=\"_blank\">Ginés Hidalgo\u003C\u002Fa> testing the \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FCMU-Perceptual-Computing-Lab\u002Fopenpose_unity_plugin\" target=\"_blank\">OpenPose Unity Plugin\u003C\u002Fa>\u003C\u002Fsup>\n\u003C\u002Fp>\n\n### Runtime Analysis\nWe show an inference time comparison between the 3 available pose estimation libraries (same hardware and conditions): OpenPose, Alpha-Pose (fast Pytorch version), and Mask R-CNN. The OpenPose runtime is constant, while the runtime of Alpha-Pose and Mask R-CNN grow linearly with the number of people. More details [**here**](https:\u002F\u002Farxiv.org\u002Fabs\u002F1812.08008).\n\n\u003Cp align=\"center\">\n    \u003Cimg src=\".github\u002Fmedia\u002Fopenpose_vs_competition.png\" width=\"360\">\n\u003C\u002Fp>\n\n\n\n## Features\n**Main Functionality**:\n- **2D real-time multi-person keypoint detection**:\n    - 15, 18 or **25-keypoint body\u002Ffoot keypoint estimation**, including **6 foot keypoints**. **Runtime invariant to number of detected people**.\n    - **2x21-keypoint hand keypoint estimation**. **Runtime depends on number of detected people**. See [**OpenPose Training**](https:\u002F\u002Fgithub.com\u002FCMU-Perceptual-Computing-Lab\u002Fopenpose_train) for a runtime invariant alternative.\n    - **70-keypoint face keypoint estimation**. **Runtime depends on number of detected people**. See [**OpenPose Training**](https:\u002F\u002Fgithub.com\u002FCMU-Perceptual-Computing-Lab\u002Fopenpose_train) for a runtime invariant alternative.\n- [**3D real-time single-person keypoint detection**](doc\u002Fadvanced\u002F3d_reconstruction_module.md):\n    - 3D triangulation from multiple single views.\n    - Synchronization of Flir cameras handled.\n    - Compatible with Flir\u002FPoint Grey cameras.\n- [**Calibration toolbox**](doc\u002Fadvanced\u002Fcalibration_module.md): Estimation of distortion, intrinsic, and extrinsic camera parameters.\n- **Single-person tracking** for further speedup or visual smoothing.\n\n**Input**: Image, video, webcam, Flir\u002FPoint Grey, IP camera, and support to add your own custom input source (e.g., depth camera).\n\n**Output**: Basic image + keypoint display\u002Fsaving (PNG, JPG, AVI, ...), keypoint saving (JSON, XML, YML, ...), keypoints as array class, and support to add your own custom output code (e.g., some fancy UI).\n\n**OS**: Ubuntu (20, 18, 16, 14), Windows (10, 8), Mac OSX, Nvidia TX2.\n\n**Hardware compatibility**: CUDA (Nvidia GPU), OpenCL (AMD GPU), and non-GPU (CPU-only) versions.\n\n**Usage Alternatives**:\n- [**Command-line demo**](doc\u002F01_demo.md) for built-in functionality.\n- [**C++ API**](doc\u002F04_cpp_api.md\u002F) and [**Python API**](doc\u002F03_python_api.md) for custom functionality. E.g., adding your custom inputs, pre-processing, post-posprocessing, and output steps.\n\nFor further details, check the [major released features](doc\u002F07_major_released_features.md) and [release notes](doc\u002F08_release_notes.md) docs.\n\n\n\n## Related Work\n- [**OpenPose training code**](https:\u002F\u002Fgithub.com\u002FCMU-Perceptual-Computing-Lab\u002Fopenpose_train)\n- [**OpenPose foot dataset**](https:\u002F\u002Fcmu-perceptual-computing-lab.github.io\u002Ffoot_keypoint_dataset\u002F)\n- [**OpenPose Unity Plugin**](https:\u002F\u002Fgithub.com\u002FCMU-Perceptual-Computing-Lab\u002Fopenpose_unity_plugin)\n- OpenPose papers published in **IEEE TPAMI and CVPR**. Cite them in your publications if OpenPose helps your research! (Links and more details in the [Citation](#citation) section below).\n\n\n\n## Installation\nIf you want to use OpenPose without installing or writing any code, simply [download and use the latest Windows portable version of OpenPose](doc\u002Finstallation\u002F0_index.md#windows-portable-demo)!\n\nOtherwise, you could [build OpenPose from source](doc\u002Finstallation\u002F0_index.md#compiling-and-running-openpose-from-source). See the [installation doc](doc\u002Finstallation\u002F0_index.md) for all the alternatives.\n\n\n\n## Quick Start Overview\nSimply use the OpenPose Demo from your favorite command-line tool (e.g., Windows PowerShell or Ubuntu Terminal). E.g., this example runs OpenPose on your webcam and displays the body keypoints:\n```\n# Ubuntu\n.\u002Fbuild\u002Fexamples\u002Fopenpose\u002Fopenpose.bin\n```\n```\n:: Windows - Portable Demo\nbin\\OpenPoseDemo.exe --video examples\\media\\video.avi\n```\n\nYou can also add any of the available flags in any order. E.g., the following example runs on a video (`--video {PATH}`), enables face (`--face`) and hands (`--hand`), and saves the output keypoints on JSON files on disk (`--write_json {PATH}`).\n```\n# Ubuntu\n.\u002Fbuild\u002Fexamples\u002Fopenpose\u002Fopenpose.bin --video examples\u002Fmedia\u002Fvideo.avi --face --hand --write_json output_json_folder\u002F\n```\n```\n:: Windows - Portable Demo\nbin\\OpenPoseDemo.exe --video examples\\media\\video.avi --face --hand --write_json output_json_folder\u002F\n```\n\nOptionally, you can also extend OpenPose's functionality from its Python and C++ APIs. After [installing](doc\u002Finstallation\u002F0_index.md) OpenPose, check its [official doc](doc\u002F00_index.md) for a quick overview of all the alternatives and tutorials.\n\n\n\n## Send Us Feedback!\nOur library is open source for research purposes, and we want to improve it! So let us know (create a new GitHub issue or pull request, email us, etc.) if you...\n1. Find\u002Ffix any bug (in functionality or speed) or know how to speed up or improve any part of OpenPose.\n2. Want to add\u002Fshow some cool functionality\u002Fdemo\u002Fproject made on top of OpenPose. We can add your project link to our [Community-based Projects](doc\u002F10_community_projects.md) section or even integrate it with OpenPose!\n\n\n\n## Citation\nPlease cite these papers in your publications if OpenPose helps your research. All of OpenPose is based on [OpenPose: Realtime Multi-Person 2D Pose Estimation using Part Affinity Fields](https:\u002F\u002Farxiv.org\u002Fabs\u002F1812.08008), while the hand and face detectors also use [Hand Keypoint Detection in Single Images using Multiview Bootstrapping](https:\u002F\u002Farxiv.org\u002Fabs\u002F1704.07809) (the face detector was trained using the same procedure as the hand detector).\n\n    @article{8765346,\n      author = {Z. {Cao} and G. {Hidalgo Martinez} and T. {Simon} and S. {Wei} and Y. A. {Sheikh}},\n      journal = {IEEE Transactions on Pattern Analysis and Machine Intelligence},\n      title = {OpenPose: Realtime Multi-Person 2D Pose Estimation using Part Affinity Fields},\n      year = {2019}\n    }\n\n    @inproceedings{simon2017hand,\n      author = {Tomas Simon and Hanbyul Joo and Iain Matthews and Yaser Sheikh},\n      booktitle = {CVPR},\n      title = {Hand Keypoint Detection in Single Images using Multiview Bootstrapping},\n      year = {2017}\n    }\n\n    @inproceedings{cao2017realtime,\n      author = {Zhe Cao and Tomas Simon and Shih-En Wei and Yaser Sheikh},\n      booktitle = {CVPR},\n      title = {Realtime Multi-Person 2D Pose Estimation using Part Affinity Fields},\n      year = {2017}\n    }\n\n    @inproceedings{wei2016cpm,\n      author = {Shih-En Wei and Varun Ramakrishna and Takeo Kanade and Yaser Sheikh},\n      booktitle = {CVPR},\n      title = {Convolutional pose machines},\n      year = {2016}\n    }\n\nPaper links:\n- OpenPose: Realtime Multi-Person 2D Pose Estimation using Part Affinity Fields:\n    - [IEEE TPAMI](https:\u002F\u002Fieeexplore.ieee.org\u002Fdocument\u002F8765346)\n    - [ArXiv](https:\u002F\u002Farxiv.org\u002Fabs\u002F1812.08008)\n- [Hand Keypoint Detection in Single Images using Multiview Bootstrapping](https:\u002F\u002Farxiv.org\u002Fabs\u002F1704.07809)\n- [Realtime Multi-Person 2D Pose Estimation using Part Affinity Fields](https:\u002F\u002Farxiv.org\u002Fabs\u002F1611.08050)\n- [Convolutional Pose Machines](https:\u002F\u002Farxiv.org\u002Fabs\u002F1602.00134)\n\n\n\n## License\nOpenPose is freely available for free non-commercial use, and may be redistributed under these conditions. Please, see the [license](.\u002FLICENSE) for further details. Interested in a commercial license? Check this [FlintBox link](https:\u002F\u002Fcmu.flintbox.com\u002F#technologies\u002Fb820c21d-8443-4aa2-a49f-8919d93a8740). For commercial queries, use the `Contact` section from the [FlintBox link](https:\u002F\u002Fcmu.flintbox.com\u002F#technologies\u002Fb820c21d-8443-4aa2-a49f-8919d93a8740) and also send a copy of that message to [Yaser Sheikh](mailto:yaser@cs.cmu.edu).\n","\u003Cdiv align=\"center\">\n    \u003Cimg src=\".github\u002FLogo_main_black.png\" width=\"300\">\n\u003C\u002Fdiv>\n\n-----------------\n\n| **构建类型**   |`Linux`           |`MacOS`           |`Windows`         |\n| :---:            | :---:            | :---:            | :---:            |\n| **构建状态** | [![状态](https:\u002F\u002Fgithub.com\u002FCMU-Perceptual-Computing-Lab\u002Fopenpose\u002Fworkflows\u002FCI\u002Fbadge.svg)](https:\u002F\u002Fgithub.com\u002FCMU-Perceptual-Computing-Lab\u002Fopenpose\u002Factions) | [![状态](https:\u002F\u002Fgithub.com\u002FCMU-Perceptual-Computing-Lab\u002Fopenpose\u002Fworkflows\u002FCI\u002Fbadge.svg)](https:\u002F\u002Fgithub.com\u002FCMU-Perceptual-Computing-Lab\u002Fopenpose\u002Factions) | [![状态](https:\u002F\u002Fci.appveyor.com\u002Fapi\u002Fprojects\u002Fstatus\u002F5leescxxdwen77kg\u002Fbranch\u002Fmaster?svg=true)](https:\u002F\u002Fci.appveyor.com\u002Fproject\u002Fgineshidalgo99\u002Fopenpose\u002Fbranch\u002Fmaster) |\n\n[**OpenPose**](https:\u002F\u002Fgithub.com\u002FCMU-Perceptual-Computing-Lab\u002Fopenpose) 是**首个能够在单张图像上同时检测人体、手部、面部和脚部关键点（共计135个关键点）的实时多人系统**。\n\n该系统由[**Ginés Hidalgo**](https:\u002F\u002Fwww.gineshidalgo.com)、[**Zhe Cao**](https:\u002F\u002Fpeople.eecs.berkeley.edu\u002F~zhecao)、[**Tomas Simon**](http:\u002F\u002Fwww.cs.cmu.edu\u002F~tsimon)、[**Shih-En Wei**](https:\u002F\u002Fscholar.google.com\u002Fcitations?user=sFQD3k4AAAAJ&hl=en)、[**Yaadhav Raaj**](https:\u002F\u002Fwww.raaj.tech)、[**Hanbyul Joo**](https:\u002F\u002Fjhugestar.github.io)以及[**Yaser Sheikh**](http:\u002F\u002Fwww.cs.cmu.edu\u002F~yaser)共同开发。目前由[**Ginés Hidalgo**](https:\u002F\u002Fwww.gineshidalgo.com)和[**Yaadhav Raaj**](https:\u002F\u002Fwww.raaj.tech)维护。OpenPose 的实现离不开[**CMU 全景工作室数据集**](http:\u002F\u002Fdomedb.perception.cs.cmu.edu)。我们还要感谢所有以各种方式帮助过 OpenPose 的人，详情请参阅 [doc\u002F09_authors_and_contributors.md] 文档。\n\n\n\n\u003Cp align=\"center\">\n    \u003Cimg src=\".github\u002Fmedia\u002Fpose_face_hands.gif\" width=\"480\">\n    \u003Cbr>\n    \u003Csup>作者 Ginés Hidalgo（左）和 Hanbyul Joo（右）在 CMU 全景工作室前合影\u003C\u002Fsup>\n\u003C\u002Fp>\n\n\n\n## 目录\n1. [结果](#results)\n2. [特性](#features)\n3. [相关工作](#related-work)\n4. [安装](#installation)\n5. [快速入门概览](#quick-start-overview)\n6. [给我们反馈！](#send-us-feedback)\n7. [引用](#citation)\n8. [许可证](#license)\n\n\n\n## 结果\n### 全身（身体、脚、面部和手部）2D 姿态估计\n\u003Cp align=\"center\">\n    \u003Cimg src=\".github\u002Fmedia\u002Fdance_foot.gif\" width=\"300\">\n    \u003Cimg src=\".github\u002Fmedia\u002Fpose_face.gif\" width=\"300\">\n    \u003Cimg src=\".github\u002Fmedia\u002Fpose_hands.gif\" width=\"300\">\n    \u003Cbr>\n    \u003Csup>测试 OpenPose：（左）来自 YouTube 视频 \u003Ca href=\"https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=2DiQUX11YaY\" target=\"_blank\">悉尼疯狂 Uptown Funk 快闪舞\u003C\u002Fa> 的片段。（中、右）作者 Ginés Hidalgo 和 Tomas Simon 正在测试面部与手部关键点\u003C\u002Fsup>\n\u003C\u002Fp>\n\n### 全身 3D 姿态重建与估计\n\u003Cp align=\"center\">\n    \u003Cimg src=\".github\u002Fmedia\u002Fopenpose3d.gif\" width=\"360\">\n    \u003Cbr>\n    \u003Csup>\u003Ca href=\"https:\u002F\u002Fziutinyat.github.io\u002F\" target=\"_blank\">Tianyi Zhao\u003C\u002Fa> 正在测试 OpenPose 3D 模块\u003C\u002Fsup>\n\u003C\u002Fp>\n\n### Unity 插件\n\u003Cp align=\"center\">\n    \u003Cimg src=\".github\u002Fmedia\u002Funity_main.png\" width=\"300\">\n    \u003Cimg src=\".github\u002Fmedia\u002Funity_body_foot.png\" width=\"300\">\n    \u003Cimg src=\".github\u002Fmedia\u002Funity_hand_face.png\" width=\"300\">\n    \u003Cbr>\n    \u003Csup>\u003Ca href=\"https:\u002F\u002Fziutinyat.github.io\u002F\" target=\"_blank\">Tianyi Zhao\u003C\u002Fa> 和 \u003Ca href=\"https:\u002F\u002Fwww.gineshidalgo.com\" target=\"_blank\">Ginés Hidalgo\u003C\u002Fa> 正在测试 \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FCMU-Perceptual-Computing-Lab\u002Fopenpose_unity_plugin\" target=\"_blank\">OpenPose Unity 插件\u003C\u002Fa>\u003C\u002Fsup>\n\u003C\u002Fp>\n\n### 运行时分析\n我们对比了三种可用的姿态估计库在同一硬件和条件下进行推理的时间：OpenPose、Alpha-Pose（快速 PyTorch 版本）和 Mask R-CNN。结果显示，OpenPose 的运行时间保持恒定，而 Alpha-Pose 和 Mask R-CNN 的运行时间则会随着人数增加呈线性增长。更多细节请参见 [此处](https:\u002F\u002Farxiv.org\u002Fabs\u002F1812.08008)。\n\n\u003Cp align=\"center\">\n    \u003Cimg src=\".github\u002Fmedia\u002Fopenpose_vs_competition.png\" width=\"360\">\n\u003C\u002Fp>\n\n\n\n## 特性\n**主要功能**：\n- **2D 实时多人关键点检测**：\n    - 可进行 15、18 或 **25 点身体\u002F脚部关键点估计**，其中包括 **6 个脚部关键点**。**运行时间与检测到的人数无关**。\n    - **2×21 点手部关键点估计**。**运行时间取决于检测到的人数**。如需运行时间不受人数影响的替代方案，请参阅 [**OpenPose 训练代码**](https:\u002F\u002Fgithub.com\u002FCMU-Perceptual-Computing-Lab\u002Fopenpose_train)。\n    - **70 点面部关键点估计**。**运行时间取决于检测到的人数**。如需运行时间不受人数影响的替代方案，请参阅 [**OpenPose 训练代码**](https:\u002F\u002Fgithub.com\u002FCMU-Perceptual-Computing-Lab\u002Fopenpose_train)。\n- [**3D 实时单人关键点检测**](doc\u002Fadvanced\u002F3d_reconstruction_module.md)：\n    - 从多个单视图进行 3D 三角测量。\n    - 自动处理 Flir 摄像头的同步问题。\n    - 兼容 Flir\u002FPoint Grey 摄像头。\n- [**标定工具箱**](doc\u002Fadvanced\u002Fcalibration_module.md)：用于估计相机的畸变、内参和外参。\n- **单人跟踪**功能，可进一步提升速度或实现平滑的视觉效果。\n\n**输入**：图像、视频、网络摄像头、Flir\u002FPoint Grey 摄像头、IP 摄像头，还支持添加自定义输入源（例如深度相机）。\n\n**输出**：基础图像 + 关键点显示\u002F保存（PNG、JPG、AVI 等），关键点保存（JSON、XML、YML 等），关键点以数组形式输出，同时支持用户自定义输出代码（例如精美的 UI）。\n\n**操作系统**：Ubuntu（20、18、16、14）、Windows（10、8）、Mac OSX、Nvidia TX2。\n\n**硬件兼容性**：CUDA（Nvidia GPU）、OpenCL（AMD GPU）以及无 GPU 的 CPU 专用版本。\n\n**使用方式**：\n- 使用 [**命令行演示**](doc\u002F01_demo.md) 来体验内置功能。\n- 使用 [**C++ API**](doc\u002F04_cpp_api.md\u002F) 和 [**Python API**](doc\u002F03_python_api.md) 来实现自定义功能，例如添加自定义输入、预处理、后处理及输出步骤。\n\n更多详细信息，请参阅 [重大发布特性](doc\u002F07_major_released_features.md) 和 [发布说明](doc\u002F08_release_notes.md) 文档。\n\n\n\n## 相关工作\n- [**OpenPose 训练代码**](https:\u002F\u002Fgithub.com\u002FCMU-Perceptual-Computing-Lab\u002Fopenpose_train)\n- [**OpenPose 脚部关键点数据集**](https:\u002F\u002Fcmu-perceptual-computing-lab.github.io\u002Ffoot_keypoint_dataset\u002F)\n- [**OpenPose Unity 插件**](https:\u002F\u002Fgithub.com\u002FCMU-Perceptual-Computing-Lab\u002Fopenpose_unity_plugin)\n- OpenPose 的相关论文已发表于 **IEEE TPAMI 和 CVPR** 期刊。若您的研究受到 OpenPose 的启发，请在您的论文中引用这些文献！（链接及更多详情请参见下方的 [引用](#citation) 部分）。\n\n## 安装\n如果您想在不安装或编写任何代码的情况下使用 OpenPose，只需[下载并使用最新的 Windows 便携版 OpenPose](doc\u002Finstallation\u002F0_index.md#windows-portable-demo)即可！\n\n否则，您也可以[从源代码构建 OpenPose](doc\u002Finstallation\u002F0_index.md#compiling-and-running-openpose-from-source)。有关所有替代方案，请参阅[安装文档](doc\u002Finstallation\u002F0_index.md)。\n\n\n\n## 快速入门概览\n只需在您喜欢的命令行工具中（例如 Windows PowerShell 或 Ubuntu 终端）使用 OpenPose 演示程序即可。例如，以下示例将在您的网络摄像头上运行 OpenPose，并显示人体关键点：\n```\n# Ubuntu\n.\u002Fbuild\u002Fexamples\u002Fopenpose\u002Fopenpose.bin\n```\n```\n:: Windows - 便携版演示\nbin\\OpenPoseDemo.exe --video examples\\media\\video.avi\n```\n\n您还可以按任意顺序添加任何可用的标志。例如，以下示例将处理视频文件（`--video {PATH}`），启用人脸检测（`--face`）和手部检测（`--hand`），并将输出的关键点保存为 JSON 文件到磁盘（`--write_json {PATH}`）。\n```\n# Ubuntu\n.\u002Fbuild\u002Fexamples\u002Fopenpose\u002Fopenpose.bin --video examples\u002Fmedia\u002Fvideo.avi --face --hand --write_json output_json_folder\u002F\n```\n```\n:: Windows - 便携版演示\nbin\\OpenPoseDemo.exe --video examples\\media\\video.avi --face --hand --write_json output_json_folder\u002F\n```\n\n此外，您还可以通过 Python 和 C++ API 扩展 OpenPose 的功能。在[安装](doc\u002Finstallation\u002F0_index.md) OpenPose 后，请查阅其[官方文档](doc\u002F00_index.md)，以快速了解所有替代方案和教程。\n\n\n\n## 向我们提供反馈！\n我们的库是面向研究用途的开源项目，我们希望不断改进它！因此，如果您……请随时告诉我们（创建新的 GitHub 问题或拉取请求、发送电子邮件等）：\n1. 发现或修复了任何错误（无论是功能上的还是速度上的），或者知道如何提升 OpenPose 某一部分的速度或性能。\n2. 希望添加或展示基于 OpenPose 构建的酷炫功能、演示或项目。我们可以将您的项目链接添加到我们的[社区项目](doc\u002F10_community_projects.md)部分，甚至将其集成到 OpenPose 中！\n\n\n\n## 引用\n如果 OpenPose 对您的研究有所帮助，请在您的出版物中引用以下论文。OpenPose 的核心基于[OpenPose：基于部位亲和场的实时多人二维姿态估计](https:\u002F\u002Farxiv.org\u002Fabs\u002F1812.08008)，而手部和面部检测器还参考了[利用多视角自举法进行单张图像中的手部关键点检测](https:\u002F\u002Farxiv.org\u002Fabs\u002F1704.07809)的研究成果（面部检测器的训练过程与手部检测器相同）。\n\n    @article{8765346,\n      author = {Z. {Cao} and G. {Hidalgo Martinez} and T. {Simon} and S. {Wei} and Y. A. {Sheikh}},\n      journal = {IEEE Transactions on Pattern Analysis and Machine Intelligence},\n      title = {OpenPose: Realtime Multi-Person 2D Pose Estimation using Part Affinity Fields},\n      year = {2019}\n    }\n\n    @inproceedings{simon2017hand,\n      author = {Tomas Simon and Hanbyul Joo and Iain Matthews and Yaser Sheikh},\n      booktitle = {CVPR},\n      title = {Hand Keypoint Detection in Single Images using Multiview Bootstrapping},\n      year = {2017}\n    }\n\n    @inproceedings{cao2017realtime,\n      author = {Zhe Cao and Tomas Simon and Shih-En Wei and Yaser Sheikh},\n      booktitle = {CVPR},\n      title = {Realtime Multi-Person 2D Pose Estimation using Part Affinity Fields},\n      year = {2017}\n    }\n\n    @inproceedings{wei2016cpm,\n      author = {Shih-En Wei and Varun Ramakrishna and Takeo Kanade and Yaser Sheikh},\n      booktitle = {CVPR},\n      title = {Convolutional Pose Machines},\n      year = {2016}\n    }\n\n论文链接：\n- OpenPose：实时多人二维姿态估计（基于部位亲和场）：\n    - [IEEE TPAMI](https:\u002F\u002Fieeexplore.ieee.org\u002Fdocument\u002F8765346)\n    - [ArXiv](https:\u002F\u002Farxiv.org\u002Fabs\u002F1812.08008)\n- [单张图像中的手部关键点检测（利用多视角自举法）](https:\u002F\u002Farxiv.org\u002Fabs\u002F1704.07809)\n- [实时多人二维姿态估计（基于部位亲和场）](https:\u002F\u002Farxiv.org\u002Fabs\u002F1611.08050)\n- [卷积姿态机器](https:\u002F\u002Farxiv.org\u002Fabs\u002F1602.00134)\n\n\n\n## 许可证\nOpenPose 可供免费的非商业用途使用，并可在满足特定条件的情况下重新分发。更多详细信息，请参阅[许可证](.\u002FLICENSE)。有意获取商业许可？请访问此[FlintBox 链接](https:\u002F\u002Fcmu.flintbox.com\u002F#technologies\u002Fb820c21d-8443-4aa2-a49f-8919d93a8740)。如需商业合作相关咨询，请使用该[FlintBox 链接](https:\u002F\u002Fcmu.flintbox.com\u002F#technologies\u002Fb820c21d-8443-4aa2-a49f-8919d93a8740)中的“联系”部分，并同时将相关信息抄送至[Yaser Sheikh](mailto:yaser@cs.cmu.edu)。","# OpenPose 快速上手指南\n\nOpenPose 是首个实时多人姿态估计系统，能够同时检测人体、手部、面部和脚部关键点（共 135 个关键点）。本指南将帮助您快速在本地环境中运行 OpenPose。\n\n## 环境准备\n\n### 系统要求\nOpenPose 支持以下操作系统：\n- **Linux**: Ubuntu (20.04, 18.04, 16.04, 14.04)\n- **Windows**: 10, 8\n- **macOS**: OSX\n- **嵌入式**: Nvidia TX2\n\n### 硬件兼容性\n- **GPU 加速**: 支持 CUDA (Nvidia GPU) 或 OpenCL (AMD GPU)。推荐使用 Nvidia GPU 以获得最佳性能。\n- **CPU 模式**: 支持无 GPU 环境（仅 CPU），但推理速度较慢。\n\n### 前置依赖\n若选择从源码编译，需预先安装以下依赖（以 Ubuntu 为例）：\n- CMake\n- GCC \u002F G++\n- Git\n- CUDA Toolkit (如需 GPU 加速)\n- cuDNN (如需 GPU 加速)\n- OpenCV\n- Protobuf\n- Eigen\n- GFlags & GLog\n\n> **提示**：如果您不想配置环境和编译代码，Windows 用户可直接下载官方提供的 **Portable Demo (便携版)**，解压即可运行，无需安装任何依赖。\n\n## 安装步骤\n\n您可以选择**使用便携版**（推荐新手）或**从源码编译**。\n\n### 方案 A：Windows 便携版（最简单）\n1. 访问 OpenPose 发布页面或文档中提供的链接，下载最新的 `openpose-windows-portable-demo.zip`。\n2. 解压压缩包到任意目录（路径中尽量避免包含中文或空格）。\n3. 进入解压后的文件夹，即可在 `bin` 目录下找到可执行文件。\n\n### 方案 B：从源码编译 (Linux\u002FWindows\u002FmacOS)\n以下为 Linux (Ubuntu) 下的通用编译流程：\n\n1. **克隆仓库**\n   ```bash\n   git clone https:\u002F\u002Fgithub.com\u002FCMU-Perceptual-Computing-Lab\u002Fopenpose.git\n   cd openpose\n   ```\n\n2. **安装依赖**\n   ```bash\n   sudo apt-get update\n   sudo apt-get install -y libprotobuf-dev libleveldb-dev libsnappy-dev libopencv-dev libhdf5-serial-dev protobuf-compiler\n   sudo apt-get install -y --no-install-recommends libboost-all-dev\n   sudo apt-get install -y libatlas-base-dev libflags-generalpp-dev cmake\n   # 若使用 GPU，请确保已安装 CUDA 和 cuDNN\n   ```\n\n3. **配置与编译**\n   ```bash\n   mkdir build && cd build\n   cmake ..\n   make -j`nproc`\n   ```\n   *注：Windows 用户可使用 CMake GUI 生成 Visual Studio 解决方案后进行编译；macOS 用户需确保 Xcode Command Line Tools 已安装。*\n\n## 基本使用\n\n安装完成后，您可以通过命令行直接运行演示程序。\n\n### 1. 启动摄像头实时检测（人体关键点）\n这是最简单的用法，将调用默认摄像头并实时显示人体骨架。\n\n**Linux (源码编译版):**\n```bash\n.\u002Fbuild\u002Fexamples\u002Fopenpose\u002Fopenpose.bin\n```\n\n**Windows (便携版):**\n打开命令行工具（如 PowerShell 或 CMD），进入 `bin` 目录后运行：\n```cmd\nbin\\OpenPoseDemo.exe\n```\n\n### 2. 处理视频文件并检测全身（含手部和面部）\n以下示例展示如何读取本地视频，开启面部和手部检测，并将结果保存为 JSON 文件。\n\n**Linux:**\n```bash\n.\u002Fbuild\u002Fexamples\u002Fopenpose\u002Fopenpose.bin --video examples\u002Fmedia\u002Fvideo.avi --face --hand --write_json output_json_folder\u002F\n```\n\n**Windows:**\n```cmd\nbin\\OpenPoseDemo.exe --video examples\\media\\video.avi --face --hand --write_json output_json_folder\u002F\n```\n\n### 常用参数说明\n- `--video {PATH}`: 指定输入视频路径。若不填则默认使用摄像头。\n- `--face`: 启用面部关键点检测。\n- `--hand`: 启用手部关键点检测。\n- `--write_json {PATH}`: 将检测到的关键点坐标保存为 JSON 格式到指定文件夹。\n- `--display`: 控制是否显示可视化窗口（设为 `false` 可在服务器后台运行）。\n\n更多高级功能（如 3D 重建、自定义输入源、Python\u002FC++ API 调用）请参考官方文档 `doc\u002F00_index.md`。","某智能健身创业团队正在开发一款基于普通摄像头的家庭 AI 私教应用，需要实时分析用户的深蹲、瑜伽等动作规范性。\n\n### 没有 openpose 时\n- **多部位检测割裂**：团队需分别集成三套独立算法来识别身体、手掌和面部，导致代码耦合度高且维护困难。\n- **多人场景失效**：当家庭成员同时入镜锻炼时，传统单人物模型无法区分个体，关键点数据严重混淆。\n- **延迟过高影响体验**：由于串行处理多个模型，视频流分析延迟超过 500 毫秒，用户做完动作后系统才给出反馈，毫无实时互动感。\n- **手部细节丢失**：现有方案仅能粗略定位手腕，无法捕捉手指关节状态，导致无法判断用户是否握拳或张开手掌等精细动作。\n\n### 使用 openpose 后\n- **全身一体化检测**：openpose 单次推理即可同步输出身体、手脚及面部共 135 个关键点，大幅简化了系统架构。\n- **精准的多人体追踪**：即使在三人同框的复杂场景下，openpose 也能自动为每个用户分配独立的骨骼数据，互不干扰。\n- **真正的实时反馈**：凭借高效的并行计算能力，openpose 将处理延迟压缩至毫秒级，实现了“动作即纠正”的流畅交互。\n- **精细动作可量化**：借助高精度的手部与足部估计，系统现在能准确识别脚趾抓地或手指伸展程度，让专业瑜伽指导成为可能。\n\nopenpose 通过单一模型解决全身多点实时追踪难题，让低成本摄像头也能具备专业级的动作捕捉能力。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FCMU-Perceptual-Computing-Lab_openpose_f7565e56.png","CMU-Perceptual-Computing-Lab","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002FCMU-Perceptual-Computing-Lab_70c7bbd1.png",null,"https:\u002F\u002Fgithub.com\u002FCMU-Perceptual-Computing-Lab",[80,84,88,92,96,100,103],{"name":81,"color":82,"percentage":83},"C++","#f34b7d",88.9,{"name":85,"color":86,"percentage":87},"Cuda","#3A4E3A",5.6,{"name":89,"color":90,"percentage":91},"CMake","#DA3434",3.8,{"name":93,"color":94,"percentage":95},"Shell","#89e051",1.7,{"name":97,"color":98,"percentage":99},"HCL","#844FBA",0.1,{"name":101,"color":102,"percentage":99},"Batchfile","#C1F12E",{"name":104,"color":105,"percentage":106},"Python","#3572A5",0,33922,8054,"2026-04-05T06:48:59","NOASSERTION",4,"Linux (Ubuntu 14\u002F16\u002F18\u002F20), Windows (8\u002F10), macOS, Nvidia TX2","非必需。支持 CUDA (Nvidia GPU)、OpenCL (AMD GPU) 或纯 CPU 版本。具体显存大小和 CUDA 版本未在文档中明确说明，需参考详细安装文档。","未说明",{"notes":116,"python":117,"dependencies":118},"该项目主要基于 C++ 构建，同时提供 Python 和 C++ API。支持多种输入源（图片、视频、摄像头、IP 摄像头等）和输出格式。对于 Windows 用户，官方提供了无需编译的便携版（Portable Demo）。详细的依赖库版本和编译步骤需查阅项目链接中的安装文档（installation doc）。","未说明 (提供 C++ 和 Python API，但具体版本要求需在安装文档中查询)",[89,119,120,121,122,123,124,125,126,127],"CUDA\u002FOpenCL (可选)","OpenCV","Boost","GFlags","GLog","Protobuf","HDF5","LMDB","Caffe (内置或外部依赖)",[14,13],[67,130,131,132,133,134,135,136,137,138,139,140,141,142,143,144,145,146,147,148],"computer-vision","machine-learning","cpp","caffe","opencv","human-pose-estimation","real-time","deep-learning","human-behavior-understanding","cvpr-2017","multi-person","foot-estimation","keypoints","face","pose","pose-estimation","human-pose","keypoint-detection","hand-estimation","2026-03-27T02:49:30.150509","2026-04-06T05:16:38.029850",[152,157,162,167,172,176],{"id":153,"question_zh":154,"answer_zh":155,"source_url":156},17217,"如何在 Google Colab 上安装和运行 OpenPose？","可以使用以下脚本在 GPU 启用的 Colab 运行时中构建并运行 OpenPose。如果遇到 CUDA\u002FcuDNN 版本兼容性问题，可以尝试运行 `!apt install --allow-change-held-packages libcudnn8=8.1.0.77-1+cuda11.2` 来安装兼容版本。\n\n安装脚本示例：\n```bash\n! apt update\n! apt install -y cmake sudo libopencv-dev\n! git clone https:\u002F\u002Fgithub.com\u002FCMU-Perceptual-Computing-Lab\u002Fopenpose.git\n! cd openpose\u002Fubuntu && .\u002Finstall_cmake.sh && .\u002Finstall_cuda.sh && .\u002Finstall_cudnn.sh\n! cd openpose && git pull origin master && rm -r build || true && mkdir build && cd build && cmake .. && make -j`nproc`\n```\n\n运行示例：\n```bash\n!cd openpose && .\u002Fbuild\u002Fexamples\u002Fopenpose\u002Fopenpose.bin --video examples\u002Fmedia\u002Fvideo.avi --write_json output\u002F --display 0 --render_pose 0\n```","https:\u002F\u002Fgithub.com\u002FCMU-Perceptual-Computing-Lab\u002Fopenpose\u002Fissues\u002F949",{"id":158,"question_zh":159,"answer_zh":160,"source_url":161},17218,"Python API 报错 'ImportError: cannot import name pyopenpose' 或 'DLL load failed' 如何解决？","该错误通常是因为 Python 无法找到编译后的动态链接库文件。解决方法如下：\n\n1. **Windows 用户**：不要使用过时的 `os.environ['PATH']` 方法，改为在代码开头添加以下行以指定 DLL 路径：\n```python\nimport os\nos.add_dll_directory(\"your_path\\openpose\\build\\x64\\Release\")\nos.add_dll_directory(\"your_path\\openpose\\build\\bin\")\n```\n\n2. **Linux (Ubuntu) 用户**：\n   - 找到构建生成的文件（例如：`build\u002Fpython\u002Fopenpose\u002Fpyopenpose.cpython-36m-x86_64-linux-gnu.so`）。\n   - 将其复制到 Python 包目录（如 `\u002Fusr\u002Flocal\u002Flib\u002Fpython3.6\u002Fdist-packages`）。\n   - 创建符号链接：`sudo ln -s pyopenpose.cpython-36m-x86_64-linux-gnu.so pyopenpose`。\n   - 确保 `LD_LIBRARY_PATH` 环境变量包含该目录。\n   - 修改导入语句：将 `from openpose import pyopenpose as op` 改为 `import pyopenpose as op`。","https:\u002F\u002Fgithub.com\u002FCMU-Perceptual-Computing-Lab\u002Fopenpose\u002Fissues\u002F1027",{"id":163,"question_zh":164,"answer_zh":165,"source_url":166},17219,"在 Windows 上使用 RTX 20XX 系列显卡和 CUDA 10 时遇到 'invalid device function' 错误怎么办？","这是由于 CUDA 编译器与显卡架构不匹配导致的。解决方法是在 CMake 配置中手动指定主机编译器路径。\n\n具体步骤：\n1. 打开 CMake GUI。\n2. 找到 `CUDA_HOST_COMPILER` 选项。\n3. 将其默认值修改为你的 Visual Studio 编译器路径，例如：\n`C:\\Program Files (x86)\\Microsoft Visual Studio\\2017\\Community\\VC\\Tools\\MSVC\\14.16.27023\\bin\\HostX86\\x64`\n（请根据实际安装的 VS 版本调整路径）。\n4. 重新生成项目并编译。\n此外，确保在 CMake 中正确设置了 x64 架构。","https:\u002F\u002Fgithub.com\u002FCMU-Perceptual-Computing-Lab\u002Fopenpose\u002Fissues\u002F996",{"id":168,"question_zh":169,"answer_zh":170,"source_url":171},17220,"运行模型下载脚本 (getModels.bat) 失败或无法获取模型文件怎么办？","如果官方模型下载地址不可用或下载失败，可以尝试以下方案：\n1. 检查网络连接，有时服务器负载过高会导致暂时无法访问。\n2. 参考社区提供的替代下载源或镜像站点（如部分用户提到的中文技术博客中提供的网盘链接）。\n3. 手动下载模型文件并放置到 OpenPose 的 `models` 目录下。\n注意：确保下载的模型版本与你的 OpenPose 版本兼容。如果问题持续，建议升级到最新版本的 OpenPose，因为新版本可能修复了下载脚本的问题。","https:\u002F\u002Fgithub.com\u002FCMU-Perceptual-Computing-Lab\u002Fopenpose\u002Fissues\u002F1602",{"id":173,"question_zh":174,"answer_zh":175,"source_url":161},17221,"如何在 Jupyter Notebook 或本地环境中验证 pyopenpose 是否安装成功？","可以通过以下简单命令验证模块是否正确加载：\n\n在 Jupyter Notebook 或 Python Shell 中执行：\n```python\nimport openpose\nprint(dir(openpose))\n```\n如果输出列表中包含 `pyopenpose`，则说明安装成功。如果没有列出，说明 `.so` (Linux) 或 `.dll` (Windows) 文件路径配置不正确，或者文件未复制到 Python 能够识别的路径中。",{"id":177,"question_zh":178,"answer_zh":179,"source_url":161},17222,"OpenPose 运行时出现 CUDA 显存不足 (Out of Memory) 错误如何解决？","可以通过降低网络分辨率来减少显存占用。在运行 OpenPose 的命令或 Python 脚本参数中，修改 `--net_resolution` 参数。\n\n例如，在 Python API 的自定义参数字典中添加或修改：\n```python\nparams = {\n    \"model_folder\": \"\u002Fpath\u002Fto\u002Fmodels\",\n    \"net_resolution\": \"368x368\",  # 默认可能更高，降低此值可节省显存\n    # 其他参数...\n}\n```\n或者在命令行中使用：\n`--net_resolution 368x368`\n根据显存大小选择合适的分辨率，直到不再报错。",[181,185,189,193,197,201,205,209,213,217,221,225,229,233,237],{"id":182,"version":183,"summary_zh":77,"released_at":184},99412,"v1.7.0","2020-11-17T05:48:13",{"id":186,"version":187,"summary_zh":77,"released_at":188},99413,"v1.6.0","2020-04-27T03:50:43",{"id":190,"version":191,"summary_zh":77,"released_at":192},99414,"v1.5.1","2019-09-04T01:40:28",{"id":194,"version":195,"summary_zh":77,"released_at":196},99415,"v1.5.0","2019-05-16T13:14:12",{"id":198,"version":199,"summary_zh":77,"released_at":200},99416,"v1.4.0","2018-09-02T02:42:57",{"id":202,"version":203,"summary_zh":77,"released_at":204},99417,"v1.3.0","2018-03-24T16:31:23",{"id":206,"version":207,"summary_zh":77,"released_at":208},99418,"v1.2.1","2018-01-09T20:20:39",{"id":210,"version":211,"summary_zh":77,"released_at":212},99419,"v1.2.0","2017-11-03T22:01:42",{"id":214,"version":215,"summary_zh":77,"released_at":216},99420,"v1.1.0","2017-10-31T21:55:59",{"id":218,"version":219,"summary_zh":77,"released_at":220},99421,"v1.0.2","2017-10-31T21:54:44",{"id":222,"version":223,"summary_zh":77,"released_at":224},99422,"v1.0.1","2017-10-31T21:53:36",{"id":226,"version":227,"summary_zh":77,"released_at":228},99423,"v1.0.0","2017-10-31T21:49:39",{"id":230,"version":231,"summary_zh":77,"released_at":232},99424,"v1.0.0-rc3","2017-10-31T22:03:24",{"id":234,"version":235,"summary_zh":77,"released_at":236},99425,"v1.0.0-rc2","2017-10-31T21:50:08",{"id":238,"version":239,"summary_zh":77,"released_at":240},99426,"v1.0.0-rc1","2017-10-31T21:50:18"]