[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"similar-Tencent--TFace":3,"tool-Tencent--TFace":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":78,"owner_location":78,"owner_email":78,"owner_twitter":78,"owner_website":79,"owner_url":80,"languages":81,"stars":98,"forks":99,"last_commit_at":100,"license":101,"difficulty_score":10,"env_os":102,"env_gpu":102,"env_ram":102,"env_deps":103,"category_tags":106,"github_topics":78,"view_count":10,"oss_zip_url":78,"oss_zip_packed_at":78,"status":16,"created_at":107,"updated_at":108,"faqs":109,"releases":139},568,"Tencent\u002FTFace","TFace","A trusty face analysis research platform developed by Tencent Youtu Lab","TFace 是腾讯优图实验室打造的一款值得信赖的人脸分析研究平台。它提供高性能的分布式训练框架，并开源了多种高效算法的实现代码，致力于帮助社区复现和跟进前沿技术。\n\nTFace 涵盖人脸识别、人脸安全、人脸质量及面部属性四大核心模块。针对实际应用场景中的难点，它实现了多项顶尖算法，包括基于频率域的隐私保护识别、DeepFake 视频检测以及人脸防伪技术。这些成果多次发表于 CVPR、ICCV、ECCV 等顶级学术会议，充分展现了其在学术与工业界的领先实力。\n\n无论是探索人脸领域最新进展的研究人员，还是需要集成先进算法的开发者，都能从 TFace 中获益。平台内置的隐私保护方案和课程学习损失函数等技术亮点，有效降低了复现门槛，为构建更安全、精准的人脸应用系统提供了坚实基础。","\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FTencent_TFace_readme_50c78ac57d34.png\" title=\"Logo\" width=\"600\" \u002F>\n\n## Introduction\n\nTFace: A trusty face analysis research platform developed by Tencent Youtu Lab. It provides a high-performance distributed training framework and releases our efficient methods implementations.\nSome of the algorithms are self-developed, and we believe the released codes benefits researchers to follow.\n\nThis project consists of several modules: **Face Recognition**, **Face Security**, **Face Quality** and **Facial Attribute**.\n\n### Face Recognition\nThis module implements various state-of-art algorithms for face recognition.\n\n#### Paper List:\n**`2025.02`**: `UIFace: Unleashing Inherent Model Capabilities to Enhance Intra-Class Diversity in Synthetic Face Recognition` accpted by **ICLR2025**.[[paper](https:\u002F\u002Fopenreview.net\u002Fforum?id=riieAeQBJm)]\n\n**`2024.12`**: `SlerpFace: Face Template Protection via Spherical Linear Interpolation` accpted by **AAAI2025**.[[paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2407.03043)]\n\n**`2024.03`**: `Privacy-Preserving Face Recognition Using Trainable Feature Subtraction` accpted by **CVPR2024**.[[paper](https:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent\u002FCVPR2024\u002Fpapers\u002FMi_Privacy-Preserving_Face_Recognition_Using_Trainable_Feature_Subtraction_CVPR_2024_paper.pdf)]\n\n**`2023.10`**: `Privacy-Preserving Face Recognition Using Random Frequency Components` accpted by **ICCV2023**.[[paper](https:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent\u002FICCV2023\u002Fpapers\u002FMi_Privacy-Preserving_Face_Recognition_Using_Random_Frequency_Components_ICCV_2023_paper.pdf)]\n\n**`2022.9`**: `Privacy-Preserving Face Recognition with Learnable Privacy Budgets in Frequency Domain` accepted by **ECCV2022**. \n[[paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2207.07316)]\n\n**`2022.9`**: `DuetFace: Collaborative Privacy-Preserving Face Recognition via Channel Splitting in the Frequency Domain` accepted by **ACMMM2022**. [[paper](https:\u002F\u002Fdl.acm.org\u002Fdoi\u002Fabs\u002F10.1145\u002F3503161.3548303)]\n\n**`2022.6`**: `Evaluation-oriented knowledge distillation for deep face recognition` accepted by **CVPR2022**. [[paper](https:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent\u002FCVPR2022\u002Fpapers\u002FHuang_Evaluation-Oriented_Knowledge_Distillation_for_Deep_Face_Recognition_CVPR_2022_paper.pdf)]\n\n**`2021.3`**: `Consistent Instance False Positive Improves Fairness in Face Recognition` accepted by **CVPR2021**. [[paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2106.05519)]\n\n**`2021.3`**: `Spherical Confidence Learning for Face Recognition` accepted by **CVPR2021**. [[paper](https:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent\u002FCVPR2021\u002Fpapers\u002FLi_Spherical_Confidence_Learning_for_Face_Recognition_CVPR_2021_paper.pdf)] \n\n**`2020.8`**: `Improving Face Recognition from Hard Samples via Distribution Distillation Loss` accepted by **ECCV2020**. [[paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2002.03662)]\n\n**`2020.3`**: `Curricularface: adaptive curriculum learning loss for deep face recognition` has been accepted by **CVPR2020**. [[paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2004.00288)]\n\n### Face Security\nThis module implements various state-of-art algorithms for face security.\n\n#### Paper List:\n\n**`2023.09`**:  `Sibling-Attack: Rethinking Transferable Adversarial Attacks against Face Recognition` accepted by **CVPR2023**\n\n**`2021.12`**:  `Dual Contrastive Learning for General Face Forgery Detection` accepted by **AAAI2022**\n\n**`2021.12`**:  `Exploiting Fine-grained Face Forgery Clues via Progressive Enhancement Learning` accepted by **AAAI2022**\n\n**`2021.12`**:  `Delving into the Local: Dynamic Inconsistency Learning for DeepFake Video Detection` accepted by **AAAI2022**\n\n**`2021.12`**:  `Feature Generation and Hypothesis Verification for Reliable Face Anti-Spoofing` accepted by **AAAI2022**\n\n**`2021.07`**: `Spatiotemporal Inconsistency Learning for DeepFake Video Detection`  accepted by **ACM MM2021**[[paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2109.01860.pdf)] [[Analysis](https:\u002F\u002Fmp.weixin.qq.com\u002Fs\u002FUMzXD4cpK4q9GXK76dbeww)]\n\n**`2021.07`**: `Adaptive Normalized Representation Learning for Generalizable Face Anti-Spoofing`  accepted by **ACM MM2021**[[paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2108.02667)]\n\n**`2021.07`**: `Structure Destruction and Content Combination for Face Anti-Spoofing`  accepted by **IJCB2021**[[paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2107.10628)]\n\n**`2021.04`**: `Adv-Makeup: A New Imperceptible and Transferable Attack on Face Recognition`  accepted by **IJCAI2021**[[paper](https:\u002F\u002Fwww.ijcai.org\u002Fproceedings\u002F2021\u002F0173.pdf)]\n\n**`2021.04`**: `Dual Reweighting Domain Generalization for Face Presentation Attack Detection`  accepted by **IJCAI2021**[[paper](https:\u002F\u002Fwww.ijcai.org\u002Fproceedings\u002F2021\u002F0120.pdf)]\n\n**`2021.03`**: `Delving into Data: Effectively Substitute Training for Black-box Attack` accepted by **CVPR2021**. [[paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2106.05519)]\n\n**`2020.12`**: `Generalizable Representation Learning for Mixture Domain Face Anti-Spoofing` accepted by **AAAI2021**. [[paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2105.02453)]\n\n**`2020.12`**: `Local Relation Learning for Face Forgery Detection` accepted by **AAAI2021**. [[paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2105.02577)]\n\n**`2020.06`**: `Face Anti-Spoofing via Disentangled Representation Learning` accepted by **ECCV2020**. [[paper](https:\u002F\u002Fwww.ecva.net\u002Fpapers\u002Feccv_2020\u002Fpapers_ECCV\u002Fpapers\u002F123640630.pdf)]\n\n### Face Quality\n\nThis module implements the SDD-FIQA algorithm for face quality.\n\n#### Paper List:\n\n**`2021.3`**: `SDD-FIQA: Unsupervised Face Image Quality Assessment with Similarity Distribution Distance` accepted by **CVPR2021**. [[paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2103.05977)]\n\n### Facial Attribute\n\nThis module implements the M3DFEL algorithm for facial attribute.\n\n#### Paper List:\n\n**`2023.6`**: ` Rethinking the Learning Paradigm for Dynamic Facial Expression Recognition` accepted by **CVPR2023**. [[paper](https:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent\u002FCVPR2023\u002Fpapers\u002FWang_Rethinking_the_Learning_Paradigm_for_Dynamic_Facial_Expression_Recognition_CVPR_2023_paper.pdf)]\n","\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FTencent_TFace_readme_50c78ac57d34.png\" title=\"Logo\" width=\"600\" \u002F>\n\n## 介绍\n\nTFace：由腾讯优图实验室开发的可靠人脸分析研究平台。它提供了一个高性能的分布式训练框架，并发布了我们的高效方法实现。部分算法为自主研发，我们相信发布的代码有助于研究人员跟进研究。\n\n本项目包含多个模块：**人脸识别**、**人脸安全**、**人脸质量**和**面部属性**。\n\n### 人脸识别\n本模块实现了各种用于人脸识别的最先进算法（State-of-the-Art）。\n\n#### 论文列表：\n**`2025.02`**: `UIFace: Unleashing Inherent Model Capabilities to Enhance Intra-Class Diversity in Synthetic Face Recognition` 被 **ICLR2025** 录用。[[论文](https:\u002F\u002Fopenreview.net\u002Fforum?id=riieAeQBJm)]\n\n**`2024.12`**: `SlerpFace: Face Template Protection via Spherical Linear Interpolation` 被 **AAAI2025** 录用。[[论文](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2407.03043)]\n\n**`2024.03`**: `Privacy-Preserving Face Recognition Using Trainable Feature Subtraction` 被 **CVPR2024** 录用。[[论文](https:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent\u002FCVPR2024\u002Fpapers\u002FMi_Privacy-Preserving_Face_Recognition_Using_Trainable_Feature_Subtraction_CVPR_2024_paper.pdf)]\n\n**`2023.10`**: `Privacy-Preserving Face Recognition Using Random Frequency Components` 被 **ICCV2023** 录用。[[论文](https:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent\u002FICCV2023\u002Fpapers\u002FMi_Privacy-Preserving_Face_Recognition_Using_Random_Frequency_Components_ICCV_2023_paper.pdf)]\n\n**`2022.9`**: `Privacy-Preserving Face Recognition with Learnable Privacy Budgets in Frequency Domain` 被 **ECCV2022** 录用。 \n[[论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2207.07316)]\n\n**`2022.9`**: `DuetFace: Collaborative Privacy-Preserving Face Recognition via Channel Splitting in the Frequency Domain` 被 **ACMMM2022** 录用。[[论文](https:\u002F\u002Fdl.acm.org\u002Fdoi\u002Fabs\u002F10.1145\u002F3503161.3548303)]\n\n**`2022.6`**: `Evaluation-oriented knowledge distillation for deep face recognition` 被 **CVPR2022** 录用。[[论文](https:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent\u002FCVPR2022\u002Fpapers\u002FHuang_Evaluation-Oriented_Knowledge_Distillation_for_Deep_Face_Recognition_CVPR_2022_paper.pdf)]\n\n**`2021.3`**: `Consistent Instance False Positive Improves Fairness in Face Recognition` 被 **CVPR2021** 录用。[[论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2106.05519)]\n\n**`2021.3`**: `Spherical Confidence Learning for Face Recognition` 被 **CVPR2021** 录用。[[论文](https:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent\u002FCVPR2021\u002Fpapers\u002FLi_Spherical_Confidence_Learning_for_Face_Recognition_CVPR_2021_paper.pdf)] \n\n**`2020.8`**: `Improving Face Recognition from Hard Samples via Distribution Distillation Loss` 被 **ECCV2020** 录用。[[论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2002.03662)]\n\n**`2020.3`**: `Curricularface: adaptive curriculum learning loss for deep face recognition` 已被 **CVPR2020** 录用。[[论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2004.00288)]\n\n### 人脸安全\n本模块实现了各种用于人脸安全的最先进算法。\n\n#### 论文列表：\n\n**`2023.09`**:  `Sibling-Attack: Rethinking Transferable Adversarial Attacks against Face Recognition` 被 **CVPR2023** 录用\n\n**`2021.12`**:  `Dual Contrastive Learning for General Face Forgery Detection` 被 **AAAI2022** 录用\n\n**`2021.12`**:  `Exploiting Fine-grained Face Forgery Clues via Progressive Enhancement Learning` 被 **AAAI2022** 录用\n\n**`2021.12`**:  `Delving into the Local: Dynamic Inconsistency Learning for DeepFake Video Detection` 被 **AAAI2022** 录用\n\n**`2021.12`**:  `Feature Generation and Hypothesis Verification for Reliable Face Anti-Spoofing` 被 **AAAI2022** 录用\n\n**`2021.07`**: `Spatiotemporal Inconsistency Learning for DeepFake Video Detection` 被 **ACM MM2021** 录用 [[论文](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2109.01860.pdf)] [[分析](https:\u002F\u002Fmp.weixin.qq.com\u002Fs\u002FUMzXD4cpK4q9GXK76dbeww)]\n\n**`2021.07`**: `Adaptive Normalized Representation Learning for Generalizable Face Anti-Spoofing` 被 **ACM MM2021** 录用 [[论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2108.02667)]\n\n**`2021.07`**: `Structure Destruction and Content Combination for Face Anti-Spoofing` 被 **IJCB2021** 录用 [[论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2107.10628)]\n\n**`2021.04`**: `Adv-Makeup: A New Imperceptible and Transferable Attack on Face Recognition` 被 **IJCAI2021** 录用 [[论文](https:\u002F\u002Fwww.ijcai.org\u002Fproceedings\u002F2021\u002F0173.pdf)]\n\n**`2021.04`**: `Dual Reweighting Domain Generalization for Face Presentation Attack Detection` 被 **IJCAI2021** 录用 [[论文](https:\u002F\u002Fwww.ijcai.org\u002Fproceedings\u002F2021\u002F0120.pdf)]\n\n**`2021.03`**: `Delving into Data: Effectively Substitute Training for Black-box Attack` 被 **CVPR2021** 录用。[[论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2106.05519)]\n\n**`2020.12`**: `Generalizable Representation Learning for Mixture Domain Face Anti-Spoofing` 被 **AAAI2021** 录用。[[论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2105.02453)]\n\n**`2020.12`**: `Local Relation Learning for Face Forgery Detection` 被 **AAAI2021** 录用。[[论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2105.02577)]\n\n**`2020.06`**: `Face Anti-Spoofing via Disentangled Representation Learning` 被 **ECCV2020** 录用。[[论文](https:\u002F\u002Fwww.ecva.net\u002Fpapers\u002Feccv_2020\u002Fpapers_ECCV\u002Fpapers\u002F123640630.pdf)]\n\n### 人脸质量\n\n本模块实现了用于人脸质量的 SDD-FIQA 算法。\n\n#### 论文列表：\n\n**`2021.3`**: `SDD-FIQA: Unsupervised Face Image Quality Assessment with Similarity Distribution Distance` 被 **CVPR2021** 录用。[[论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2103.05977)]\n\n### 面部属性\n\n本模块实现了用于面部属性的 M3DFEL 算法。\n\n#### 论文列表：\n\n**`2023.6`**: ` Rethinking the Learning Paradigm for Dynamic Facial Expression Recognition` 被 **CVPR2023** 录用。[[论文](https:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent\u002FCVPR2023\u002Fpapers\u002FWang_Rethinking_the_Learning_Paradigm_for_Dynamic_Facial_Expression_Recognition_CVPR_2023_paper.pdf)]","# TFace 快速上手指南\n\n## 简介\nTFace 是由腾讯优图实验室（Tencent Youtu Lab）开发的可信赖人脸分析研究平台。它提供了一个高性能的分布式训练框架，并发布了高效的算法实现。部分算法为自研，旨在帮助研究人员跟进前沿技术。\n\n项目主要包含以下四大模块：\n*   **人脸识别 (Face Recognition)**\n*   **人脸安全 (Face Security)**\n*   **人脸质量 (Face Quality)**\n*   **面部属性 (Facial Attribute)**\n\n---\n\n## 环境准备\n由于本项目涉及深度学习模型训练与推理，建议配置如下环境：\n\n*   **操作系统**: Linux (推荐 Ubuntu 18.04+) 或 Windows\n*   **编程语言**: Python 3.8+\n*   **深度学习框架**: PyTorch (根据具体模块需求)\n*   **硬件要求**: NVIDIA GPU (CUDA 支持)\n*   **依赖管理**: `pip` 或 `conda`\n\n> **提示**: 国内开发者建议使用国内镜像源以加速依赖包下载。\n\n---\n\n## 安装步骤\n\n### 1. 克隆代码库\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002Ftencent\u002FTFace.git\ncd TFace\n```\n\n### 2. 安装依赖\n根据项目根目录下的 `requirements.txt` 安装所需依赖：\n```bash\npip install -r requirements.txt\n```\n*(若遇到网络问题，可添加国内镜像参数)*\n```bash\npip install -r requirements.txt -i https:\u002F\u002Fpypi.tuna.tsinghua.edu.cn\u002Fsimple\n```\n\n### 3. 验证安装\n确保相关深度学习库已正确加载：\n```bash\npython -c \"import torch; print(torch.__version__)\"\n```\n\n---\n\n## 基本使用\n\nTFace 提供了模块化设计，用户可根据需求调用不同功能模块。以下是各模块的核心功能概览：\n\n### 1. 人脸识别 (Face Recognition)\n实现了多种先进的人脸识别算法，支持高内聚类多样性及隐私保护。\n*   **典型应用**: 身份验证、人脸检索。\n*   **参考算法**: Curricularface, UIFace, SlerpFace 等。\n\n### 2. 人脸安全 (Face Security)\n专注于对抗攻击防御、活体检测及 DeepFake 视频检测。\n*   **典型应用**: 防伪检测、隐私保护、恶意攻击防御。\n*   **参考算法**: Dual Contrastive Learning, Sibling-Attack 防御方案等。\n\n### 3. 人脸质量 (Face Quality)\n提供无监督图像质量评估算法。\n*   **典型应用**: 筛选低质量人脸图片，提升后续识别准确率。\n*   **参考算法**: SDD-FIQA。\n\n### 4. 面部属性 (Facial Attribute)\n实现动态面部表情识别及相关属性分析。\n*   **典型应用**: 表情分析、情绪识别。\n*   **参考算法**: M3DFEL。\n\n---\n\n## 学术支持与更新\n本项目持续跟进顶级会议研究成果，核心算法已在多个权威会议发表，包括但不限于：\n\n*   **ICLR 2025**: `UIFace` (增强合成人脸识别中的类内多样性)\n*   **AAAI 2025**: `SlerpFace` (基于球面线性插值的模板保护)\n*   **CVPR 2024**: `Privacy-Preserving Face Recognition Using Trainable Feature Subtraction`\n*   **ICCV 2023**: `Privacy-Preserving Face Recognition Using Random Frequency Components`\n*   **ECCV 2022 \u002F CVPR 2022 \u002F CVPR 2021**: 多项关于知识蒸馏、公平性及隐私保护的算法。\n\n更多详细论文及技术细节，请参考官方仓库中的 `doc` 目录或对应链接。","某互联网金融公司正在构建新一代线上身份认证系统，核心需求是既要保证高并发下的人脸识别精度，又要严防 DeepFake 伪造攻击。\n\n### 没有 TFace 时\n- 研发团队需从零开始设计网络结构，训练周期长达数月且收敛困难，难以达到业界顶尖水平。\n- 面对日益复杂的活体检测手段，自研方案容易漏过高清合成视频，存在严重的资金盗刷风险。\n- 缺乏成熟的隐私计算模块，直接上传原始人脸特征存在法律合规风险，可能违反数据安全法规。\n- 分布式训练环境配置繁琐，导致多卡并行效率低下，算力成本高昂且维护难度大。\n\n### 使用 TFace 后\n- 直接复用 Curricularface 等顶会论文代码，大幅缩短模型研发周期并显著提升识别准确率。\n- 内置双对比学习与动态不一致性学习模块，有效拦截各类 DeepFake 换脸及伪造行为。\n- 集成频率域隐私保护算法，在不影响识别率的前提下满足数据脱敏合规要求，降低法律风险。\n- 依托其高性能分布式训练框架，轻松实现大规模数据集的快速迭代与部署，优化了硬件资源利用率。\n\nTFace 通过提供经过验证的 SOTA 算法与工程化框架，帮助团队将重心从底层基建转移至业务创新。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FTencent_TFace_50c78ac5.png","Tencent","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002FTencent_f7e55588.png","",null,"https:\u002F\u002Fopensource.tencent.com","https:\u002F\u002Fgithub.com\u002FTencent",[82,86,90,94],{"name":83,"color":84,"percentage":85},"Python","#3572A5",98.2,{"name":87,"color":88,"percentage":89},"C++","#f34b7d",1.2,{"name":91,"color":92,"percentage":93},"Shell","#89e051",0.4,{"name":95,"color":96,"percentage":97},"CMake","#DA3434",0.1,1504,245,"2026-03-30T04:02:22","Apache-2.0","未说明",{"notes":104,"python":102,"dependencies":105},"提供的 README 内容仅包含项目简介及论文列表，未包含具体的运行环境配置、依赖库版本及硬件要求说明。",[102],[14,13],"2026-03-27T02:49:30.150509","2026-04-06T05:16:57.349042",[110,115,120,124,129,134],{"id":111,"question_zh":112,"answer_zh":113,"source_url":114},2313,"CIFP 论文中提到的超参数 `ru+` 在代码中是如何定义的？","代码中并未显式定义该变量。其实现逻辑是通过仅对正 Logit 应用惩罚来实现的（见 `cos_theta.scatter_(1, label.view(-1, 1).long(), target_cos_theta_m)`）。从数学推导上看，分子分母同时除以该项后，两者是等价的。","https:\u002F\u002Fgithub.com\u002FTencent\u002FTFace\u002Fissues\u002F7",{"id":116,"question_zh":117,"answer_zh":118,"source_url":119},2314,"训练过程中 Loss 突然变为 NaN 如何解决？","这通常与学习率过高或数据集噪声较大有关。建议尝试调小 Margin（尤其是 VGG 等噪声较大的数据集），并降低 Learning Rate。有用户反馈在内存限制导致 Batch Size 较小（如 64）时，降低学习率后可以恢复正常训练。","https:\u002F\u002Fgithub.com\u002FTencent\u002FTFace\u002Fissues\u002F109",{"id":121,"question_zh":122,"answer_zh":123,"source_url":119},2315,"DCTDP 项目如何划分训练\u002F测试集以及进行模型评估？","测试集仍使用标准验证集（包括 LFW、CFP、IJBB、IJBC 等）。测试脚本需参考训练代码自行改写。评估方法通常是利用在 VGGFace2 上训练好的 Backbone 作为特征提取器生成 Embedding，直接在测试集上计算 Verification Accuracy，论文中的准确率即指此指标，无需在 LFW 上微调。",{"id":125,"question_zh":126,"answer_zh":127,"source_url":128},2316,"SDD-FIQA 模型输出的质量分数高低范围是多少？","并没有严格的固定阈值来区分低、中、高质量，不同区间存在重叠。学术数据集可能缺少非人脸或误检样本，而商业版本会在训练中引入此类图像以便通过低质量分数过滤。分数范围依赖于具体模型，且受输入检测框质量影响。","https:\u002F\u002Fgithub.com\u002FTencent\u002FTFace\u002Fissues\u002F20",{"id":130,"question_zh":131,"answer_zh":132,"source_url":133},2317,"如何获取 LFW 等验证数据集的二进制文件（如 lfw.bin）？","可以通过自行构建 Pipeline 生成，例如参考 facial-recognition 仓库中的 validation_prep.md 流程。也可以在互联网上搜索已发布的 bin 文件，或有其他开源社区提供了现成的验证数据集资源链接（如 https:\u002F\u002Fgithub.com\u002FLingcan-M\u002Fverification-datasets-such-as-lfw.bin）。","https:\u002F\u002Fgithub.com\u002FTencent\u002FTFace\u002Fissues\u002F27",{"id":135,"question_zh":136,"answer_zh":137,"source_url":138},2318,"DuetFace 项目是否已经开源？","是的，DuetFace 的代码已经开源。","https:\u002F\u002Fgithub.com\u002FTencent\u002FTFace\u002Fissues\u002F77",[]]