[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"similar-kuaikuaikim--dface":3,"tool-kuaikuaikim--dface":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",140436,2,"2026-04-05T23:32:43",[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":76,"owner_avatar_url":77,"owner_bio":78,"owner_company":79,"owner_location":80,"owner_email":81,"owner_twitter":82,"owner_website":83,"owner_url":84,"languages":85,"stars":90,"forks":91,"last_commit_at":92,"license":93,"difficulty_score":10,"env_os":94,"env_gpu":95,"env_ram":96,"env_deps":97,"category_tags":104,"github_topics":105,"view_count":10,"oss_zip_url":82,"oss_zip_packed_at":82,"status":16,"created_at":111,"updated_at":112,"faqs":113,"releases":134},980,"kuaikuaikim\u002Fdface","dface","Deep learning face detection and recognition, implemented by pytorch. (pytorch实现的人脸检测和人脸识别)","dface 是一个免费开源的人脸检测与识别工具，基于 PyTorch 深度学习框架开发。它能快速准确地从图像或视频中定位人脸（检测），并识别不同人的身份（识别），解决了传统方法在复杂场景下精度低、速度慢的问题——比如在光线变化、遮挡或多人环境中也能稳定工作。开发者和研究人员最适合使用它：你可以轻松集成到安防系统、身份验证应用或学术项目中，还能基于开源代码训练自定义模型或贡献新功能。技术亮点在于它采用 MTCNN 级联神经网络架构，支持 NVIDIA GPU 加速（Linux 版本性能极佳），并利用 PyTorch 的动态计算图特性，让模型调整更灵活高效。跨平台兼容性也很强，无论是 Linux、Windows 还是 Mac 系统都能流畅运行。作为完全免费的工具，dface 降低了 AI 人脸技术的使用门槛，社区还提供 SDK 扩展防欺骗、追踪等高级功能，欢迎动手实践或参与优化！（字数：298）","\u003Cdiv align=center>\n\u003Ca href=\"https:\u002F\u002Fdface.tech\" target=\"_blank\">\u003Cimg src=\"http:\u002F\u002Fdftech.oss-cn-hangzhou.aliyuncs.com\u002Fweb\u002FDFACE-logo_dark.png\" width=\"160\">\u003C\u002Fa>\n\u003C\u002Fdiv>\n\n-----------------\n# Dface • [![License](http:\u002F\u002Fdftech.oss-cn-hangzhou.aliyuncs.com\u002Fopendface\u002Fimg\u002Fapache_2.svg)](https:\u002F\u002Fopensource.org\u002Flicenses\u002FApache-2.0)\n\n\n| **`Linux CPU`** | **`Linux GPU`** | **`Mac OS CPU`** | **`Windows CPU`** |\n|-----------------|---------------------|------------------|-------------------|\n| [![Build Status](http:\u002F\u002Fdftech.oss-cn-hangzhou.aliyuncs.com\u002Fopendface\u002Fimg\u002Fbuild_pass.svg)](http:\u002F\u002Fdftech.oss-cn-hangzhou.aliyuncs.com\u002Fopendface\u002Fimg\u002Fbuild_pass.svg) | [![Build Status](http:\u002F\u002Fdftech.oss-cn-hangzhou.aliyuncs.com\u002Fopendface\u002Fimg\u002Fbuild_pass.svg)](http:\u002F\u002Fdftech.oss-cn-hangzhou.aliyuncs.com\u002Fopendface\u002Fimg\u002Fbuild_pass.svg) | [![Build Status](http:\u002F\u002Fdftech.oss-cn-hangzhou.aliyuncs.com\u002Fopendface\u002Fimg\u002Fbuild_pass.svg)](http:\u002F\u002Fdftech.oss-cn-hangzhou.aliyuncs.com\u002Fopendface\u002Fimg\u002Fbuild_pass.svg) | [![Build Status](http:\u002F\u002Fdftech.oss-cn-hangzhou.aliyuncs.com\u002Fopendface\u002Fimg\u002Fbuild_pass.svg)](http:\u002F\u002Fdftech.oss-cn-hangzhou.aliyuncs.com\u002Fopendface\u002Fimg\u002Fbuild_pass.svg) |\n\n\n**Free and open source face detection. Based on the MTCNN**\n\n[Official Website(https:\u002F\u002Fdface.tech)](https:\u002F\u002Fdface.tech)  \n\n**We also provide fully face recognize SDK, Contains tracking, detection, face recognition, face anti-spoofing and so on. See [dface.tech](https:\u002F\u002Fdface.tech) for details.**  \n![DFACE SDK](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fkuaikuaikim_dface_readme_42c4349b611a.gif)\n\n\n**Dface** is an open source software for face detection and recognition. All features implemented by the **[pytorch](https:\u002F\u002Fgithub.com\u002Fpytorch\u002Fpytorch)** (the facebook deeplearning framework). With PyTorch, we use a technique called reverse-mode auto-differentiation, which allows developer to change the way your network behaves arbitrarily with zero lag or overhead.\nDFace inherit these advanced characteristic, that make it dynamic and ease code review.\n\nDFace support GPU acceleration with NVIDIA cuda. We highly recommend you use the linux GPU version.It's very fast and extremely realtime.\n\nOur inspiration comes from several research papers on this topic, as well as current and past work such as [Joint Face Detection and Alignment using Multi-task Cascaded Convolutional Networks](https:\u002F\u002Farxiv.org\u002Fabs\u002F1604.02878) and face recognition topic [FaceNet: A Unified Embedding for Face Recognition and Clustering](https:\u002F\u002Farxiv.org\u002Fabs\u002F1503.03832)\n\n**MTCNN Structure**　　\n\n![Pnet](http:\u002F\u002Fdftech.oss-cn-hangzhou.aliyuncs.com\u002Fopendface\u002Fimg\u002Fpnet.jpg)\n![Rnet](http:\u002F\u002Fdftech.oss-cn-hangzhou.aliyuncs.com\u002Fopendface\u002Fimg\u002Frnet.jpg)\n![Onet](http:\u002F\u002Fdftech.oss-cn-hangzhou.aliyuncs.com\u002Fopendface\u002Fimg\u002Fonet.jpg)\n\n**If you want to contribute to DFace, please review the CONTRIBUTING.md in the project.We use [Slack](https:\u002F\u002Fdfaceio.slack.com\u002F) for tracking requests and bugs. Also you can following the QQ group 681403076 or my wechat jinkuaikuai005**\n\n\n## TODO(contribute to DFace)\n- Based on cener loss or triplet loss implement the face conpare. Recommended Model is ResNet inception v2. Refer this [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F1503.03832) and [FaceNet](https:\u002F\u002Fgithub.com\u002Fdavidsandberg\u002Ffacenet)\n- Face Anti-Spoofing, distinguish from face light and texture。Recomend with the LBP algorithm and SVM.\n- 3D mask  Anti-Spoofing.\n- Mobile first with caffe2 and c++.\n- Tensor rt migration.\n- Docker support, gpu version\n\n## Installation\n\nDFace has two major module, detection and recognition.In these two, We provide all tutorials about how to train a model and running.\nFirst setting a pytorch and cv2. We suggest Anaconda to make a virtual and independent python envirment.**If you want to train on GPU,please install Nvidia cuda and cudnn.**\n\n### Requirements\n* cuda 8.0\n* anaconda\n* pytorch\n* torchvision\n* cv2\n* matplotlib  \n\n\n```shell\ngit clone https:\u002F\u002Fgithub.com\u002Fkuaikuaikim\u002Fdface.git\n```\n\n\nAlso we provide a anaconda environment dependency list called environment.yml (windows please use environment-win64.yml,Mac environment_osx.yaml) in the root path. \nYou can create your DFace environment very easily.\n```shell\ncd DFace\n\nconda env create -f path\u002Fto\u002Fenvironment.yml\n```\n\nAdd Dface to your local python path  \n\n```shell\nexport PYTHONPATH=$PYTHONPATH:{your local DFace root path}\n```\n\n\n### Face Detetion and Recognition\n\nIf you are interested in how to train a mtcnn model, you can follow next step.\n\n#### Train mtcnn Model\nMTCNN have three networks called **PNet**, **RNet** and **ONet**.So we should train it on three stage, and each stage depend on previous network which will generate train data to feed current train net, also propel the minimum loss between two networks.\nPlease download the train face **datasets** before your training. We use **[WIDER FACE](http:\u002F\u002Fmmlab.ie.cuhk.edu.hk\u002Fprojects\u002FWIDERFace\u002F)** and **[CelebA](http:\u002F\u002Fmmlab.ie.cuhk.edu.hk\u002Fprojects\u002FCelebA.html)**  .WIDER FACE is used for training face classification and face bounding box, also CelebA is used for face landmarks. The original wider face annotation file is matlab format, you must transform it to text. I have put the transformed annotation text file into [anno_store\u002Fwider_origin_anno.txt](https:\u002F\u002Fgithub.com\u002Fkuaikuaikim\u002FDFace\u002Fblob\u002Fmaster\u002Fanno_store\u002Fwider_origin_anno.txt). This file is related to the following parameter called  --anno_file.\n\n\n* Create the DFace train data temporary folder, this folder is involved in the following parameter --dface_traindata_store \n\n```shell\nmkdir {your dface traindata folder}\n```   \n\n\n* Generate PNet Train data and annotation file\n\n```shell\npython dface\u002Fprepare_data\u002Fgen_Pnet_train_data.py --prefix_path {annotation file image prefix path, just your local wider face images folder} --dface_traindata_store  {dface train data temporary folder you made before }  --anno_file ｛wider face original combined  annotation file, default anno_store\u002Fwider_origin_anno.txt}\n```\n* Assemble annotation file and shuffle it\n\n```shell\npython dface\u002Fprepare_data\u002Fassemble_pnet_imglist.py\n```\n* Train PNet model\n\n```shell\npython dface\u002Ftrain_net\u002Ftrain_p_net.py\n```\n* Generate RNet Train data and annotation file\n\n```shell\npython dface\u002Fprepare_data\u002Fgen_Rnet_train_data.py --prefix_path {annotation file image prefix path, just your local wider face images folder} --dface_traindata_store {dface train data temporary folder you made before } --anno_file ｛wider face original combined  annotation file, default anno_store\u002Fwider_origin_anno.txt} --pmodel_file {your PNet model file trained before}\n```\n* Assemble annotation file and shuffle it\n\n```shell\npython dface\u002Fprepare_data\u002Fassemble_rnet_imglist.py\n```\n* Train RNet model\n\n```shell\npython dface\u002Ftrain_net\u002Ftrain_r_net.py\n```\n* Generate ONet Train data and annotation file\n\n```shell\npython dface\u002Fprepare_data\u002Fgen_Onet_train_data.py --prefix_path {annotation file image prefix path, just your local wider face images folder} --dface_traindata_store {dface train data temporary folder you made before } --anno_file ｛wider face original combined  annotation file, default anno_store\u002Fwider_origin_anno.txt} --pmodel_file {your PNet model file trained before} --rmodel_file {your RNet model file trained before}\n```\n* Generate ONet Train landmarks data\n\n```shell\npython dface\u002Fprepare_data\u002Fgen_landmark_48.py\n```\n* Assemble annotation file and shuffle it\n\n```shell\npython dface\u002Fprepare_data\u002Fassemble_onet_imglist.py\n```\n* Train ONet model\n\n```shell\npython dface\u002Ftrain_net\u002Ftrain_o_net.py\n```\n\n#### Test face detection  \n**If you don't want to train,i have put onet_epoch.pt,pnet_epoch.pt,rnet_epoch.pt in model_store folder.You just try test_image.py**\n\n```shell\npython test_image.py\n```    \n\n\n## Demo  \n\n![mtcnn](http:\u002F\u002Fdftech.oss-cn-hangzhou.aliyuncs.com\u002Fopendface\u002Fimg\u002Fdface_demoall.PNG)  \n\n\n### QQ交流群  \n![](http:\u002F\u002Fdftech.oss-cn-hangzhou.aliyuncs.com\u002Fopendface\u002Fimg\u002Fdfaceqqsm.png)\n\n\n#### 681403076  \n\n#### 本人微信(wechat)  \n##### cobbestne\n\n\n## License  \n\n[Apache License 2.0](LICENSE)\n\n\n## Reference\n\n* [OpenFace](https:\u002F\u002Fgithub.com\u002Fcmusatyalab\u002Fopenface)\n","\u003Cdiv align=center>\n\u003Ca href=\"https:\u002F\u002Fdface.tech\" target=\"_blank\">\u003Cimg src=\"http:\u002F\u002Fdftech.oss-cn-hangzhou.aliyuncs.com\u002Fweb\u002FDFACE-logo_dark.png\" width=\"160\">\u003C\u002Fa>\n\u003C\u002Fdiv>\n\n-----------------\n# Dface • [![License](http:\u002F\u002Fdftech.oss-cn-hangzhou.aliyuncs.com\u002Fopendface\u002Fimg\u002Fapache_2.svg)](https:\u002F\u002Fopensource.org\u002Flicenses\u002FApache-2.0)\n\n\n| **`Linux CPU`** | **`Linux GPU`** | **`Mac OS CPU`** | **`Windows CPU`** |\n|-----------------|---------------------|------------------|-------------------|\n| [![Build Status](http:\u002F\u002Fdftech.oss-cn-hangzhou.aliyuncs.com\u002Fopendface\u002Fimg\u002Fbuild_pass.svg)](http:\u002F\u002Fdftech.oss-cn-hangzhou.aliyuncs.com\u002Fopendface\u002Fimg\u002Fbuild_pass.svg) | [![Build Status](http:\u002F\u002Fdftech.oss-cn-hangzhou.aliyuncs.com\u002Fopendface\u002Fimg\u002Fbuild_pass.svg)](http:\u002F\u002Fdftech.oss-cn-hangzhou.aliyuncs.com\u002Fopendface\u002Fimg\u002Fbuild_pass.svg) | [![Build Status](http:\u002F\u002Fdftech.oss-cn-hangzhou.aliyuncs.com\u002Fopendface\u002Fimg\u002Fbuild_pass.svg)](http:\u002F\u002Fdftech.oss-cn-hangzhou.aliyuncs.com\u002Fopendface\u002Fimg\u002Fbuild_pass.svg) | [![Build Status](http:\u002F\u002Fdftech.oss-cn-hangzhou.aliyuncs.com\u002Fopendface\u002Fimg\u002Fbuild_pass.svg)](http:\u002F\u002Fdftech.oss-cn-hangzhou.aliyuncs.com\u002Fopendface\u002Fimg\u002Fbuild_pass.svg) |\n\n\n**免费开源的人脸检测（face detection）工具。基于多任务级联卷积神经网络（MTCNN）**\n\n[官方网站(https:\u002F\u002Fdface.tech)](https:\u002F\u002Fdface.tech)  \n\n**我们还提供完整的人脸识别SDK，包含人脸追踪、检测、识别、活体检测（face anti-spoofing）等功能。详情请见 [dface.tech](https:\u002F\u002Fdface.tech)。**  \n![DFACE SDK](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fkuaikuaikim_dface_readme_42c4349b611a.gif)\n\n\n**Dface** 是一款开源的人脸检测与识别软件，所有功能均基于 **[PyTorch](https:\u002F\u002Fgithub.com\u002Fpytorch\u002Fpytorch)**（Facebook 深度学习框架）实现。借助 PyTorch 的反向模式自动微分（reverse-mode auto-differentiation）技术，开发者可任意调整网络行为而无需额外开销。DFace 继承了这些先进特性，使其具备动态灵活性和代码可读性。\n\nDFace 支持 NVIDIA CUDA 的 GPU 加速。我们强烈推荐使用 Linux GPU 版本，其运行速度极快且能实现超实时性能。\n\n我们的灵感源自该领域的多项研究论文及现有成果，例如 [《基于多任务级联卷积网络的联合人脸检测与对齐》](https:\u002F\u002Farxiv.org\u002Fabs\u002F1604.02878) 和人脸识别方向的 [《FaceNet：用于人脸识别与聚类的统一嵌入》](https:\u002F\u002Farxiv.org\u002Fabs\u002F1503.03832)。\n\n**MTCNN 结构**　　\n\n![Pnet](http:\u002F\u002Fdftech.oss-cn-hangzhou.aliyuncs.com\u002Fopendface\u002Fimg\u002Fpnet.jpg)\n![Rnet](http:\u002F\u002Fdftech.oss-cn-hangzhou.aliyuncs.com\u002Fopendface\u002Fimg\u002Frnet.jpg)\n![Onet](http:\u002F\u002Fdftech.oss-cn-hangzhou.aliyuncs.com\u002Fopendface\u002Fimg\u002Fonet.jpg)\n\n**如需为 DFace 贡献代码，请查阅项目中的 CONTRIBUTING.md。我们使用 [Slack](https:\u002F\u002Fdfaceio.slack.com\u002F) 跟踪需求与问题，也可加入 QQ 群 681403076 或添加微信 jinkuaikuai005 联系。**\n\n\n## 待办事项（欢迎贡献）\n- 基于中心损失（center loss）或三元组损失（triplet loss）实现人脸比对。推荐使用 ResNet Inception v2 模型。参考 [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F1503.03832) 和 [FaceNet](https:\u002F\u002Fgithub.com\u002Fdavidsandberg\u002Ffacenet)\n- 人脸活体检测（face anti-spoofing），通过人脸光照与纹理特征进行区分。推荐结合局部二值模式（LBP）算法与支持向量机（SVM）\n- 3D 面具活体检测\n- 优先适配移动端，基于 Caffe2 和 C++\n- 迁移到 TensorRT\n- 支持 Docker（含 GPU 版本）\n\n## 安装\n\nDFace 包含检测（detection）与识别（recognition）两大核心模块，我们提供了完整的模型训练与运行教程。首先配置 PyTorch 和 OpenCV (cv2)，建议使用 Anaconda 创建独立的 Python 环境。**若需 GPU 训练，请安装 NVIDIA CUDA 和 cuDNN。**\n\n### 环境要求\n* CUDA 8.0\n* Anaconda\n* PyTorch\n* torchvision\n* OpenCV (cv2)\n* matplotlib  \n\n\n```shell\ngit clone https:\u002F\u002Fgithub.com\u002Fkuaikuaikim\u002Fdface.git\n```\n\n\n项目根目录提供 Anaconda 环境依赖列表（Windows 用户使用 environment-win64.yml，Mac 用户使用 environment_osx.yaml）。您可快速创建 DFace 环境：\n```shell\ncd DFace\n\nconda env create -f path\u002Fto\u002Fenvironment.yml\n```\n\n将 Dface 添加至本地 Python 路径  \n\n```shell\nexport PYTHONPATH=$PYTHONPATH:{your local DFace root path}\n```\n\n### 人脸检测与识别（Face Detection and Recognition）\n\n如果您想了解如何训练 MTCNN（多任务级联卷积神经网络）模型，可按以下步骤操作。\n\n#### 训练 MTCNN 模型\nMTCNN 包含三个网络：**PNet**、**RNet** 和 **ONet**。因此需分三阶段训练，每阶段依赖前一阶段生成的训练数据，并推动两个网络间的最小损失。\n训练前请下载人脸 **数据集（datasets）**。我们使用 **[WIDER FACE](http:\u002F\u002Fmmlab.ie.cuhk.edu.hk\u002Fprojects\u002FWIDERFace\u002F)** 和 **[CelebA](http:\u002F\u002Fmmlab.ie.cuhk.edu.hk\u002Fprojects\u002FCelebA.html)**。WIDER FACE 用于训练人脸分类和边界框检测，CelebA 用于人脸关键点定位。原始 WIDER FACE 标注文件为 MATLAB 格式，需转换为文本格式。转换后的标注文件已存放在 [anno_store\u002Fwider_origin_anno.txt](https:\u002F\u002Fgithub.com\u002Fkuaikuaikim\u002FDFace\u002Fblob\u002Fmaster\u002Fanno_store\u002Fwider_origin_anno.txt)，该文件与后续参数 `--anno_file` 相关。\n\n* 创建 DFace 训练数据临时文件夹（对应参数 `--dface_traindata_store`）\n\n```shell\nmkdir {your dface traindata folder}\n```   \n\n* 生成 PNet 训练数据及标注文件\n\n```shell\npython dface\u002Fprepare_data\u002Fgen_Pnet_train_data.py --prefix_path {annotation file image prefix path, just your local wider face images folder} --dface_traindata_store  {dface train data temporary folder you made before }  --anno_file ｛wider face original combined  annotation file, default anno_store\u002Fwider_origin_anno.txt}\n```\n* 整合标注文件并打乱顺序\n\n```shell\npython dface\u002Fprepare_data\u002Fassemble_pnet_imglist.py\n```\n* 训练 PNet 模型\n\n```shell\npython dface\u002Ftrain_net\u002Ftrain_p_net.py\n```\n* 生成 RNet 训练数据及标注文件\n\n```shell\npython dface\u002Fprepare_data\u002Fgen_Rnet_train_data.py --prefix_path {annotation file image prefix path, just your local wider face images folder} --dface_traindata_store {dface train data temporary folder you made before } --anno_file ｛wider face original combined  annotation file, default anno_store\u002Fwider_origin_anno.txt} --pmodel_file {your PNet model file trained before}\n```\n* 整合标注文件并打乱顺序\n\n```shell\npython dface\u002Fprepare_data\u002Fassemble_rnet_imglist.py\n```\n* 训练 RNet 模型\n\n```shell\npython dface\u002Ftrain_net\u002Ftrain_r_net.py\n```\n* 生成 ONet 训练数据及标注文件\n\n```shell\npython dface\u002Fprepare_data\u002Fgen_Onet_train_data.py --prefix_path {annotation file image prefix path, just your local wider face images folder} --dface_traindata_store {dface train data temporary folder you made before } --anno_file ｛wider face original combined  annotation file, default anno_store\u002Fwider_origin_anno.txt} --pmodel_file {your PNet model file trained before} --rmodel_file {your RNet model file trained before}\n```\n* 生成 ONet 训练关键点数据\n\n```shell\npython dface\u002Fprepare_data\u002Fgen_landmark_48.py\n```\n* 整合标注文件并打乱顺序\n\n```shell\npython dface\u002Fprepare_data\u002Fassemble_onet_imglist.py\n```\n* 训练 ONet 模型\n\n```shell\npython dface\u002Ftrain_net\u002Ftrain_o_net.py\n```\n\n#### 测试人脸检测  \n**如无需训练，model_store 文件夹已提供 onet_epoch.pt、pnet_epoch.pt、rnet_epoch.pt 模型文件，可直接运行 test_image.py**\n\n```shell\npython test_image.py\n```    \n\n\n## 演示  \n\n![mtcnn](http:\u002F\u002Fdftech.oss-cn-hangzhou.aliyuncs.com\u002Fopendface\u002Fimg\u002Fdface_demoall.PNG)  \n\n\n### QQ 交流群  \n![](http:\u002F\u002Fdftech.oss-cn-hangzhou.aliyuncs.com\u002Fopendface\u002Fimg\u002Fdfaceqqsm.png)\n\n\n#### 681403076  \n\n#### 本人微信(wechat)  \n##### cobbestne\n\n\n## 许可证  \n\n[Apache License 2.0](LICENSE)\n\n\n## 参考文献\n\n* [OpenFace](https:\u002F\u002Fgithub.com\u002Fcmusatyalab\u002Fopenface)","# dface 快速上手指南\n\n## 环境准备\n- **系统要求**：支持 Linux (CPU\u002FGPU)、Mac OS (CPU)、Windows (CPU)\n- **前置依赖**：\n  - CUDA 8.0（仅 GPU 版本需要）\n  - Anaconda（推荐使用国内镜像加速安装：`conda config --add channels https:\u002F\u002Fmirrors.tuna.tsinghua.edu.cn\u002Fanaconda\u002Fpkgs\u002Ffree\u002F`）\n  - PyTorch\n  - torchvision\n  - OpenCV (cv2)\n  - matplotlib\n\n## 安装步骤\n1. 克隆仓库：\n   ```shell\n   git clone https:\u002F\u002Fgithub.com\u002Fkuaikuaikim\u002Fdface.git\n   cd dface\n   ```\n\n2. 创建 Conda 环境（根据系统选择对应文件）：\n   - **Linux\u002FMac**：\n     ```shell\n     conda env create -f environment.yml\n     ```\n   - **Windows**：\n     ```shell\n     conda env create -f environment-win64.yml\n     ```\n   - **Mac OS**：\n     ```shell\n     conda env create -f environment_osx.yaml\n     ```\n\n3. 激活环境并设置路径：\n   ```shell\n   conda activate dface  # 环境名根据实际创建情况调整\n   export PYTHONPATH=$PYTHONPATH:$(pwd)\n   ```\n\n## 基本使用\n使用预训练模型快速测试人脸检测（无需训练）：\n```shell\npython test_image.py\n```\n该命令将自动加载 `model_store` 目录中的预训练模型（`pnet_epoch.pt`、`rnet_epoch.pt`、`onet_epoch.pt`），处理示例图像并显示检测结果。运行后将在控制台输出人脸检测框和关键点坐标。","某安防科技公司为高端社区开发智能门禁系统，需在入口处实时检测并识别居民人脸，实现无感通行与安全管控。\n\n### 没有 dface 时\n- 依赖商业人脸SDK，单设备年授权费超5000元，200个门禁点项目成本激增，挤压利润空间。\n- 仅用CPU处理视频流，识别延迟达1.5秒，早晚高峰居民排队拥堵，投诉率上升30%。\n- 自研MTCNN模型需3个月训练调优，准确率仅85%，常误判导致门禁误锁或漏放。\n- Windows\u002FLinux环境需分别适配代码，部署耗时2周，且Mac测试机无法运行，调试效率低下。\n- 无活体检测功能，居民用手机照片即可骗过系统，安全隐患频发。\n\n### 使用 dface 后\n- 开源免费特性直接砍掉授权成本，200个门禁点节省100万元，资金转向用户体验优化。\n- GPU加速下识别速度压缩至0.15秒，门禁通行零等待，高峰期通行效率提升8倍。\n- 预训练MTCNN模型开箱即用，1天集成准确率达98.5%，误判率下降至0.2%。\n- 统一PyTorch代码库覆盖Linux\u002FWindows\u002FMac，部署时间缩短至2天，跨平台调试无缝衔接。\n- 内置LBP+CNN反欺骗模块，精准识别照片\u002F视频攻击，安全事件归零。\n\ndface以开源、高效、全栈式人脸处理能力，让智能门禁系统实现低成本、高可靠、秒级响应的实战价值。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fkuaikuaikim_dface_41f09119.png","kuaikuaikim","K.Kuai Jin","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002Fkuaikuaikim_76b3b26f.jpg","Implementation for Audio algorithm, Computer Vision, DeepLearning, High performance computing. Rich experience on C\u002FC++, Python, DSP, and Java.","wuqi microelectronics","Shanghai","314127900@qq.com",null,"https:\u002F\u002Fwww.zhihu.com\u002Fpeople\u002Fjinkuaikuai\u002Factivities","https:\u002F\u002Fgithub.com\u002Fkuaikuaikim",[86],{"name":87,"color":88,"percentage":89},"Python","#3572A5",100,1334,351,"2026-04-02T19:59:38","Apache-2.0","Linux, macOS, Windows","需要 NVIDIA GPU，CUDA 8.0，显存未说明","未说明",{"notes":98,"python":96,"dependencies":99},"建议使用 conda 创建虚拟环境；训练需下载 WIDER FACE 和 CelebA 数据集；GPU 版本仅支持 Linux 环境",[100,101,102,103],"pytorch","torchvision","opencv-python","matplotlib",[13,14],[106,107,108,100,109,110],"facedetection","facerecognition","deeplearning","mtcnn","mtcnn-pytorch","2026-03-27T02:49:30.150509","2026-04-06T08:52:25.382907",[114,119,124,129],{"id":115,"question_zh":116,"answer_zh":117,"source_url":118},4358,"是否有预训练模型？","是的，但需要通过邮件请求获取。请发送请求至维护者邮箱：jinkuaikuai@outlook.com。","https:\u002F\u002Fgithub.com\u002Fkuaikuaikim\u002Fdface\u002Fissues\u002F1",{"id":120,"question_zh":121,"answer_zh":122,"source_url":123},4359,"是否需要安装Anaconda环境？","不必要。维护者指出conda会安装过多依赖，建议避免使用Anaconda环境以减少冗余。","https:\u002F\u002Fgithub.com\u002Fkuaikuaikim\u002Fdface\u002Fissues\u002F4",{"id":125,"question_zh":126,"answer_zh":127,"source_url":128},4360,"gen_Onet_train_data.py中出现负维度错误如何解决？","维护者建议过滤掉导致负预测的样本图像。具体方法：在数据处理过程中，检查并移除tmph[i]或tmpw[i]小于或等于零的样本，以避免负维度错误。","https:\u002F\u002Fgithub.com\u002Fkuaikuaikim\u002Fdface\u002Fissues\u002F10",{"id":130,"question_zh":131,"answer_zh":132,"source_url":133},4361,"推荐的PyTorch、TorchVision、OpenCV和Matplotlib版本是什么？","可以使用conda安装environment.yml文件中指定的包版本，具体命令为：`conda env create -f path\u002Fto\u002Fenvironment.yml`。该文件包含项目所需的精确依赖配置。","https:\u002F\u002Fgithub.com\u002Fkuaikuaikim\u002Fdface\u002Fissues\u002F35",[135],{"id":136,"version":137,"summary_zh":138,"released_at":139},113468,"v0.5-sensible","v0.5 add these features:\r\n1. Modify readme, simplify train process\r\n2. Support windows anaconda\r\n3. Compatible with python3 ","2017-12-27T17:15:20"]