[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"similar-foamliu--Deep-Image-Matting":3,"tool-foamliu--Deep-Image-Matting":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":76,"owner_avatar_url":77,"owner_bio":78,"owner_company":79,"owner_location":80,"owner_email":81,"owner_twitter":79,"owner_website":82,"owner_url":83,"languages":84,"stars":93,"forks":94,"last_commit_at":95,"license":96,"difficulty_score":10,"env_os":97,"env_gpu":97,"env_ram":97,"env_deps":98,"category_tags":105,"github_topics":106,"view_count":10,"oss_zip_url":79,"oss_zip_packed_at":79,"status":16,"created_at":110,"updated_at":111,"faqs":112,"releases":143},835,"foamliu\u002FDeep-Image-Matting","Deep-Image-Matting","Deep Image Matting","Deep-Image-Matting 是一个基于深度学习的高精度图像抠图开源项目。它的核心功能是从普通照片中智能分离前景主体与背景，尤其擅长处理头发丝、毛绒衣物等复杂边缘细节。\n\n传统抠图方法在精细度上常遇瓶颈，容易出现锯齿或残留背景色。Deep-Image-Matting 通过神经网络学习像素级的透明度信息，生成平滑自然的 Alpha 遮罩，让主体无缝融入新背景。\n\n这个项目主要面向计算机视觉开发者、AI 研究人员以及追求专业级图像合成的设计师。使用者需要具备一定的编程基础，因为运行依赖 TensorFlow 和 Keras 等框架。项目中包含了数据预处理、模型训练及演示脚本，并提供了基于 VGG16 的预训练模型，方便快速上手体验。如果你正在探索图像分割技术或需要批量处理高质量抠图任务，这是一个值得参考的优秀代码库。","Just in case you are interested, [Deep Image Matting v2](https:\u002F\u002Fgithub.com\u002Ffoamliu\u002FDeep-Image-Matting-v2) is an upgraded version of this.\n\n# Deep Image Matting\nThis repository is to reproduce Deep Image Matting.\n\n## Dependencies\n- [NumPy](http:\u002F\u002Fdocs.scipy.org\u002Fdoc\u002Fnumpy-1.10.1\u002Fuser\u002Finstall.html)\n- [Tensorflow 1.9.0](https:\u002F\u002Fwww.tensorflow.org\u002F)\n- [Keras 2.1.6](https:\u002F\u002Fkeras.io\u002F#installation)\n- [OpenCV](https:\u002F\u002Fopencv-python-tutroals.readthedocs.io\u002Fen\u002Flatest\u002F)\n\n## Dataset\n### Adobe Deep Image Matting Dataset\nFollow the [instruction](https:\u002F\u002Fsites.google.com\u002Fview\u002Fdeepimagematting) to contact author for the dataset.\n\n### MSCOCO\nGo to [MSCOCO](http:\u002F\u002Fcocodataset.org\u002F#download) to download:\n* [2014 Train images](http:\u002F\u002Fimages.cocodataset.org\u002Fzips\u002Ftrain2014.zip)\n\n\n### PASCAL VOC\nGo to [PASCAL VOC](http:\u002F\u002Fhost.robots.ox.ac.uk\u002Fpascal\u002FVOC\u002F) to download:\n* VOC challenge 2008 [training\u002Fvalidation data](http:\u002F\u002Fhost.robots.ox.ac.uk\u002Fpascal\u002FVOC\u002Fvoc2008\u002FVOCtrainval_14-Jul-2008.tar)\n* The test data for the [VOC2008 challenge](http:\u002F\u002Fhost.robots.ox.ac.uk\u002Fpascal\u002FVOC\u002Fvoc2008\u002Findex.html#testdata)\n\n## ImageNet Pretrained Models\nDownload [VGG16](https:\u002F\u002Fgithub.com\u002Ffchollet\u002Fdeep-learning-models\u002Freleases\u002Fdownload\u002Fv0.1\u002Fvgg16_weights_tf_dim_ordering_tf_kernels.h5) into \"models\" folder.\n\n\n## Usage\n### Data Pre-processing\nExtract training images:\n```bash\n$ python pre_process.py\n```\n\n### Train\n```bash\n$ python train.py\n```\n\nIf you want to visualize during training, run in your terminal:\n```bash\n$ tensorboard --logdir path_to_current_dir\u002Flogs\n```\n\n### Demo\nDownload pre-trained Deep Image Matting [Model](https:\u002F\u002Fgithub.com\u002Ffoamliu\u002FDeep-Image-Matting\u002Freleases\u002Fdownload\u002Fv1.0\u002Ffinal.42-0.0398.hdf5) to \"models\" folder then run:\n```bash\n$ python demo.py\n```\n\nImage\u002FTrimap | Output\u002FGT | New BG\u002FCompose | \n|---|---|---|\n|![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_53fbb032b509.png)  | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_969a8dc11400.png)   | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_9da27aea6831.png) |\n|![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_c7abf304b0ed.png) | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_1f278097d59a.png) | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_809837848d41.png)|\n|![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_d18cf7fd9295.png)  | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_57e1bde26cef.png)   | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_85c8e5118a89.png) | \n|![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_f47b50ce882b.png) | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_e6a66857ca92.png) | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_a3e4d279edf0.png)|\n|![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_9a9c74ab0170.png)  | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_b472291fdb12.png)   | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_bb9ab9cb5664.png) |\n|![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_371ff8184f98.png) | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_297ef625df86.png) | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_cb95c272cb4d.png)|\n|![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_9d33a05b5f19.png)  | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_09f52f48a05f.png)   | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_ec00e72c0b92.png) |\n|![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_854c1117778f.png) | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_92502db62633.png) | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_bd0a008feae0.png)|\n|![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_45e190567764.png)  | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_7739d0a7ca09.png)   | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_3e9195d9e178.png) |\n|![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_72cc3a226021.png) | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_3f8bfa328ca4.png) | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_beef9daf4886.png)|\n|![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_a5ad0bb7b427.png)  | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_372f615ce97f.png)   | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_fef7023515fc.png) |\n|![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_c89842e6852c.png) | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_3ff4507ae74d.png) | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_67e47847ff1f.png)|\n|![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_c4a522c7a268.png)  | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_19448a33e839.png)   | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_a29ab3cd51bb.png) |\n|![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_84c6279da8ae.png) | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_642b52d909a8.png) | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_63515c3d0c03.png)|\n|![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_719ac9a62f1f.png)  | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_d5014cad228e.png)   | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_eda249cbb828.png) |\n|![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_993466513bd6.png) | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_985200a56377.png) | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_224fa6e3346e.png)|\n|![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_221960da59cd.png)  | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_bf87c689965d.png)   | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_b95561791df2.png) |\n|![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_c0ae7072647b.png) | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_f50b3bc0dc10.png) | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_79eda1e0c20c.png)|\n|![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_8b8c1817f7db.png)  | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_199ce82c5eca.png)   | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_752790e4a847.png) |\n|![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_9c55e206a874.png) | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_78fa7ee40ed4.png) | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_3fa2a8e16272.png)|\n\n","如果您感兴趣，[Deep Image Matting v2](https:\u002F\u002Fgithub.com\u002Ffoamliu\u002FDeep-Image-Matting-v2)（深度图像抠图 v2）是此项目的升级版。\n\n# Deep Image Matting（深度图像抠图）\n本仓库旨在复现 Deep Image Matting。\n\n## 依赖项\n- [NumPy](http:\u002F\u002Fdocs.scipy.org\u002Fdoc\u002Fnumpy-1.10.1\u002Fuser\u002Finstall.html)\n- [Tensorflow 1.9.0](https:\u002F\u002Fwww.tensorflow.org\u002F)\n- [Keras 2.1.6](https:\u002F\u002Fkeras.io\u002F#installation)\n- [OpenCV](https:\u002F\u002Fopencv-python-tutroals.readthedocs.io\u002Fen\u002Flatest\u002F)\n\n## 数据集\n### Adobe Deep Image Matting 数据集\n请遵循 [说明](https:\u002F\u002Fsites.google.com\u002Fview\u002Fdeepimagematting) 联系作者以获取数据集。\n\n### MSCOCO\n前往 [MSCOCO](http:\u002F\u002Fcocodataset.org\u002F#download) 下载：\n* [2014 训练图像](http:\u002F\u002Fimages.cocodataset.org\u002Fzips\u002Ftrain2014.zip)\n\n\n### PASCAL VOC\n前往 [PASCAL VOC](http:\u002F\u002Fhost.robots.ox.ac.uk\u002Fpascal\u002FVOC\u002F) 下载：\n* VOC 挑战赛 2008 [训练\u002F验证数据](http:\u002F\u002Fhost.robots.ox.ac.uk\u002Fpascal\u002FVOC\u002Fvoc2008\u002FVOCtrainval_14-Jul-2008.tar)\n* [VOC2008 挑战赛](http:\u002F\u002Fhost.robots.ox.ac.uk\u002Fpascal\u002FVOC\u002Fvoc2008\u002Findex.html#testdata) 的测试数据\n\n## ImageNet 预训练模型\n将 [VGG16](https:\u002F\u002Fgithub.com\u002Ffchollet\u002Fdeep-learning-models\u002Freleases\u002Fdownload\u002Fv0.1\u002Fvgg16_weights_tf_dim_ordering_tf_kernels.h5) 下载到 \"models\" 文件夹中。\n\n\n## 使用方法\n### 数据预处理\n提取训练图像：\n```bash\n$ python pre_process.py\n```\n\n### 训练\n```bash\n$ python train.py\n```\n\n如果您想在训练期间进行可视化，请在终端运行：\n```bash\n$ tensorboard --logdir path_to_current_dir\u002Flogs\n```\n\n### 演示\n下载预训练的 Deep Image Matting（深度图像抠图）[模型](https:\u002F\u002Fgithub.com\u002Ffoamliu\u002FDeep-Image-Matting\u002Freleases\u002Fdownload\u002Fv1.0\u002Ffinal.42-0.0398.hdf5) 到 \"models\" 文件夹，然后运行：\n```bash\n$ python demo.py\n```\n\n| 图像\u002FTrimap（修剪图）| 输出\u002FGT（真值）| 新背景\u002F合成 | \n|---|---|---|\n|![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_53fbb032b509.png)  | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_969a8dc11400.png)   | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_9da27aea6831.png) |\n|![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_c7abf304b0ed.png) | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_1f278097d59a.png) | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_809837848d41.png)|\n|![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_d18cf7fd9295.png)  | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_57e1bde26cef.png)   | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_85c8e5118a89.png) | \n|![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_f47b50ce882b.png) | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_e6a66857ca92.png) | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_a3e4d279edf0.png)|\n|![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_9a9c74ab0170.png)  | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_b472291fdb12.png)   | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_bb9ab9cb5664.png) |\n|![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_371ff8184f98.png) | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_297ef625df86.png) | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_cb95c272cb4d.png)|\n|![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_9d33a05b5f19.png)  | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_09f52f48a05f.png)   | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_ec00e72c0b92.png) |\n|![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_854c1117778f.png) | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_92502db62633.png) | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_bd0a008feae0.png)|\n|![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_45e190567764.png)  | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_7739d0a7ca09.png)   | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_3e9195d9e178.png) |\n|![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_72cc3a226021.png) | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_3f8bfa328ca4.png) | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_beef9daf4886.png)|\n|![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_a5ad0bb7b427.png)  | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_372f615ce97f.png)   | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_fef7023515fc.png) |\n|![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_c89842e6852c.png) | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_3ff4507ae74d.png) | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_67e47847ff1f.png)|\n|![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_c4a522c7a268.png)  | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_19448a33e839.png)   | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_a29ab3cd51bb.png) |\n|![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_84c6279da8ae.png) | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_642b52d909a8.png) | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_63515c3d0c03.png)|\n|![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_719ac9a62f1f.png)  | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_d5014cad228e.png)   | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_eda249cbb828.png) |\n|![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_993466513bd6.png) | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_985200a56377.png) | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_224fa6e3346e.png)|\n|![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_221960da59cd.png)  | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_bf87c689965d.png)   | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_b95561791df2.png) |\n|![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_c0ae7072647b.png) | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_f50b3bc0dc10.png) | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_79eda1e0c20c.png)|\n|![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_8b8c1817f7db.png)  | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_199ce82c5eca.png)   | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_752790e4a847.png) |\n|![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_9c55e206a874.png) | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_78fa7ee40ed4.png) | ![image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_readme_3fa2a8e16272.png)|","# Deep-Image-Matting 快速上手指南\n\n本指南旨在帮助开发者快速部署和使用 Deep Image Matting 工具进行图像抠图处理。\n\n> **注意**：[Deep Image Matting v2](https:\u002F\u002Fgithub.com\u002Ffoamliu\u002FDeep-Image-Matting-v2) 是该工具的升级版，建议优先考虑使用新版本。\n\n## 环境准备\n\n本项目基于较旧的深度学习框架版本，请确保您的开发环境满足以下要求：\n\n*   **Python**: 兼容 Python 2.7 或 3.x (建议虚拟环境隔离)\n*   **TensorFlow**: 1.9.0\n*   **Keras**: 2.1.6\n*   **其他依赖**: NumPy, OpenCV\n\n## 安装步骤\n\n### 1. 安装依赖库\n使用 pip 安装所需的 Python 包：\n```bash\n$ pip install numpy tensorflow==1.9.0 keras==2.1.6 opencv-python\n```\n\n### 2. 下载预训练模型\n下载 VGG16 权重文件并放入项目根目录下的 `models` 文件夹中：\n*   [VGG16 Weights](https:\u002F\u002Fgithub.com\u002Ffchollet\u002Fdeep-learning-models\u002Freleases\u002Fdownload\u002Fv0.1\u002Fvgg16_weights_tf_dim_ordering_tf_kernels.h5)\n\n### 3. 准备数据集\n根据需求下载相应数据集（用于训练）：\n*   **Adobe Deep Image Matting Dataset**: 需按 [官方说明](https:\u002F\u002Fsites.google.com\u002Fview\u002Fdeepimagematting) 联系作者获取。\n*   **MSCOCO**: [2014 Train images](http:\u002F\u002Fimages.cocodataset.org\u002Fzips\u002Ftrain2014.zip)\n*   **PASCAL VOC**: [VOC2008 训练\u002F验证数据](http:\u002F\u002Fhost.robots.ox.ac.uk\u002Fpascal\u002FVOC\u002Fvoc2008\u002FVOCtrainval_14-Jul-2008.tar)\n\n## 基本使用\n\n### 1. 数据预处理\n在开始训练前，先执行预处理脚本提取训练图像：\n```bash\n$ python pre_process.py\n```\n\n### 2. 模型训练\n启动训练流程：\n```bash\n$ python train.py\n```\n如需实时监控训练过程，可在终端开启 TensorBoard：\n```bash\n$ tensorboard --logdir path_to_current_dir\u002Flogs\n```\n\n### 3. 推理演示\n下载已训练好的模型文件至 `models` 文件夹，然后运行演示脚本：\n*   [最终模型链接](https:\u002F\u002Fgithub.com\u002Ffoamliu\u002FDeep-Image-Matting\u002Freleases\u002Fdownload\u002Fv1.0\u002Ffinal.42-0.0398.hdf5)\n\n```bash\n$ python demo.py\n```\n运行后将生成输入图像、Trimap、输出 Alpha 通道及合成背景等结果。","电商视觉设计师小李正在为新款珠宝系列制作线上推广素材，需要将佩戴者从复杂街景中精准抠出并合成到纯色背景上。\n\n### 没有 Deep-Image-Matting 时\n- 面对项链反光和手部细微毛发，手动钢笔工具勾勒极其耗时，容易遗漏细节。\n- 传统抠图软件难以处理半透明物体，边缘常出现白边或锯齿，影响质感。\n- 更换背景后人物与光影不协调，需花费大量时间进行二次调色和蒙版修复。\n- 面对上百张产品图，人工逐张处理导致项目交付周期被严重拉长。\n\n### 使用 Deep-Image-Matting 后\n- 通过输入简单 Trimap，Deep-Image-Matting 能自动计算精细的 Alpha 遮罩，完美保留发丝与饰品光泽。\n- 算法生成的边缘过渡平滑自然，彻底解决了半透明材质的白边问题。\n- 直接输出合成图像，前景与背景的光影融合度显著提升，减少后期修图工作量。\n- 利用脚本批量运行模型，单张图片处理时间缩短至秒级，大幅提升整体产出效率。\n\nDeep-Image-Matting 将复杂的图像分割任务转化为自动化流程，让设计师专注于创意而非重复劳动。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Ffoamliu_Deep-Image-Matting_324331ab.png","foamliu","Yang Liu","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002Ffoamliu_11ac94fc.jpg","AGI researcher",null,"Shanghai","foamliu@yeah.net","https:\u002F\u002Ffoamliu.github.io\u002F","https:\u002F\u002Fgithub.com\u002Ffoamliu",[85,89],{"name":86,"color":87,"percentage":88},"Python","#3572A5",99.9,{"name":90,"color":91,"percentage":92},"Shell","#89e051",0.1,990,258,"2026-03-29T18:29:34","MIT","未说明",{"notes":99,"python":97,"dependencies":100},"需要联系作者获取 Adobe 数据集；需手动下载 MSCOCO、PASCAL VOC 数据集及 VGG16 预训练模型到指定文件夹；训练时可使用 tensorboard 进行可视化",[101,102,103,104],"NumPy","Tensorflow 1.9.0","Keras 2.1.6","OpenCV",[14,13],[107,108,109],"computer-vision","deep-learning","matting","2026-03-27T02:49:30.150509","2026-04-06T05:36:52.203939",[113,118,123,128,133,138],{"id":114,"question_zh":115,"answer_zh":116,"source_url":117},3593,"如何下载运行所需的 Adobe Deep Matting 数据集？","维护者无法直接提供数据集下载链接。需要按照项目说明中的指引联系原作者申请。常见报错如 FileNotFoundError 提示缺少 'Adobe_Deep_Matting_Dataset.zip' 或 'training_fg_names.txt'，均需通过联系作者获取。","https:\u002F\u002Fgithub.com\u002Ffoamliu\u002FDeep-Image-Matting\u002Fissues\u002F6",{"id":119,"question_zh":120,"answer_zh":121,"source_url":122},3594,"Demo 图片中的 Trimap 是如何生成的？","官方测试数据集中已经包含了 Trimaps。训练集的 Trimaps 生成细节未公开，但有用户反馈可以使用 Mask R-CNN 生成分割掩码后再处理得到。建议优先使用官方数据集中的 Trimap 进行测试。","https:\u002F\u002Fgithub.com\u002Ffoamliu\u002FDeep-Image-Matting\u002Fissues\u002F26",{"id":124,"question_zh":125,"answer_zh":126,"source_url":127},3595,"哪里有计算 SAD、MSE 等评估指标的源代码？","可以参考 GFM 项目的评估代码：https:\u002F\u002Fgithub.com\u002FJizhiziLi\u002FGFM\u002Fblob\u002Fmaster\u002Fcore\u002Fevaluate.py。该文件包含 SAD 和 MSE 的计算逻辑，Gradient 和 Connectivity 可能需要参考其他资料或自行实现。","https:\u002F\u002Fgithub.com\u002Ffoamliu\u002FDeep-Image-Matting\u002Fissues\u002F48",{"id":129,"question_zh":130,"answer_zh":131,"source_url":132},3596,"训练过程中 GPU 利用率低且速度慢如何解决？","可能是数据预处理过慢。建议尝试数据预取（Data Prefetching）优化，或者查看 PyTorch 版本的复现代码：https:\u002F\u002Fgithub.com\u002Ffoamliu\u002FDeep-Image-Matting-PyTorch，可能性能更好。","https:\u002F\u002Fgithub.com\u002Ffoamliu\u002FDeep-Image-Matting\u002Fissues\u002F27",{"id":134,"question_zh":135,"answer_zh":136,"source_url":137},3597,"在自定义数据集上使用预训练模型出现边缘锯齿怎么办？","通常是因为 Trimap 被错误地进行了插值缩放（interpolation）。请检查 Trimap 的预处理步骤，确保没有改变其分辨率或形状。此外，也可能是自定义数据集与训练分布差异较大导致的偏差。","https:\u002F\u002Fgithub.com\u002Ffoamliu\u002FDeep-Image-Matting\u002Fissues\u002F22",{"id":139,"question_zh":140,"answer_zh":141,"source_url":142},3598,"compositional_loss 函数中为何能访问超出 batch_y 维度的通道？","这是一个潜在的实现问题。有用户发现 batch_y 实际通道数少于 loss 函数中切片访问的范围（如访问 2:5 但只有 2 通道），这可能导致 loss 计算无效。建议检查数据生成器是否完整填充了图像、前景和背景信息。","https:\u002F\u002Fgithub.com\u002Ffoamliu\u002FDeep-Image-Matting\u002Fissues\u002F9",[144],{"id":145,"version":146,"summary_zh":79,"released_at":147},103200,"v1.0","2018-07-17T03:33:13"]