[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"similar-AntonMu--TrainYourOwnYOLO":3,"tool-AntonMu--TrainYourOwnYOLO":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":79,"owner_email":80,"owner_twitter":81,"owner_website":82,"owner_url":83,"languages":84,"stars":97,"forks":98,"last_commit_at":99,"license":100,"difficulty_score":10,"env_os":101,"env_gpu":102,"env_ram":103,"env_deps":104,"category_tags":118,"github_topics":119,"view_count":139,"oss_zip_url":140,"oss_zip_packed_at":140,"status":16,"created_at":141,"updated_at":142,"faqs":143,"releases":172},179,"AntonMu\u002FTrainYourOwnYOLO","TrainYourOwnYOLO","Train a state-of-the-art yolov3 object detector from scratch!","TrainYourOwnYOLO 是一个帮助用户从零开始训练自定义 YOLOv3 目标检测模型的开源项目。它提供了一套完整的工作流程：从使用 VoTT 工具标注图像，到基于 TensorFlow 2.3 和 Keras 2.4 训练模型，再到对新图片或视频进行推理检测。该项目特别适合希望快速构建专属物体识别能力的开发者、研究人员或学生，无需从底层实现复杂算法，只需准备自己的数据集即可上手。TrainYourOwnYOLO 支持 GPU 加速训练，并提供了 Google Colab 教程，大幅降低环境配置门槛。即使在普通 CPU 上也能运行推理，每秒处理约 2 张图像。整个项目结构清晰、文档详实，是入门和实践 YOLOv3 的实用选择。","# TrainYourOwnYOLO: Building a Custom Object Detector from Scratch [![License: CC BY 4.0](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FLicense-CC%20BY%204.0-lightgrey.svg)](https:\u002F\u002Fcreativecommons.org\u002Flicenses\u002Fby\u002F4.0\u002F) [![DOI](https:\u002F\u002Fzenodo.org\u002Fbadge\u002F197467673.svg)](https:\u002F\u002Fzenodo.org\u002Fbadge\u002Flatestdoi\u002F197467673)\n\nThis repo let's you train a custom image detector using the state-of-the-art [YOLOv3](https:\u002F\u002Fpjreddie.com\u002Fdarknet\u002Fyolo\u002F) computer vision algorithm. For a short write up check out this [medium post](https:\u002F\u002Fmedium.com\u002F@muehle\u002Fhow-to-train-your-own-yolov3-detector-from-scratch-224d10e55de2). This repo works with TensorFlow 2.3 and Keras 2.4.\n\n### Before getting started:\n\n- 🍴 **fork** this repo so that you can use it as part of your own project.\n- ⭐ **star** this repo to get notifications on future improvements.\n\n### Pipeline Overview\n\nTo build and test your YOLO object detection algorithm follow the below steps:\n\n 1. [Image Annotation](\u002F1_Image_Annotation\u002F)\n\t - Install Microsoft's Visual Object Tagging Tool (VoTT)\n\t - Annotate images\n 2. [Training](\u002F2_Training\u002F)\n \t- Download pre-trained weights\n \t- Train your custom YOLO model on annotated images \n 3. [Inference](\u002F3_Inference\u002F)\n \t- Detect objects in new images and videos\n\n## Repo structure\n+ [`1_Image_Annotation`](\u002F1_Image_Annotation\u002F): Scripts and instructions on annotating images\n+ [`2_Training`](\u002F2_Training\u002F): Scripts and instructions on training your YOLOv3 model\n+ [`3_Inference`](\u002F3_Inference\u002F): Scripts and instructions on testing your trained YOLO model on new images and videos\n+ [`Data`](\u002FData\u002F): Input Data, Output Data, Model Weights and Results\n+ [`Utils`](\u002FUtils\u002F): Utility scripts used by main scripts\n\n## Getting Started\n\n### Google Colab Tutorial \u003Ca href=\"https:\u002F\u002Fcolab.research.google.com\u002Fgithub\u002FAntonMu\u002FTrainYourOwnYOLO\u002Fblob\u002Fmaster\u002FTrainYourOwnYOLO.ipynb\" target=\"_parent\">\u003Cimg src=\"https:\u002F\u002Fcolab.research.google.com\u002Fassets\u002Fcolab-badge.svg\" alt=\"Open In Colab\"\u002F>\u003C\u002Fa>\nWith Google Colab you can skip most of the set up steps and start training your own model right away. \n\n### Requisites\nThe only hard requirement is a running version of python 3.6 or 3.7. To install python 3.7 go to \n- [python.org\u002Fdownloads](https:\u002F\u002Fwww.python.org\u002Fdownloads\u002Frelease\u002Fpython-376\u002F) \n\nand follow the installation instructions. Note that this repo has only been tested with python 3.6 and python 3.7 thus it is recommened to use either `python3.6` or `python3.7`.\n\nTo speed up training, it is recommended to use a **GPU with CUDA** support. For example on [AWS](\u002F2_Training\u002FAWS\u002F) you can use a `p2.xlarge` instance (Tesla K80 GPU with 12GB memory). Inference speed on a typical CPU is approximately ~2 images per second. If you want to use your own machine, follow the instructions at [tensorflow.org\u002Finstall\u002Fgpu](https:\u002F\u002Fwww.tensorflow.org\u002Finstall\u002Fgpu) to install CUDA drivers. Make sure to install the [correct version of CUDA and cuDNN](https:\u002F\u002Fwww.tensorflow.org\u002Finstall\u002Fsource#linux). \n\n\n### Installation\n\n#### Setting up Virtual Environment [Linux or Mac]\n\nClone this repo with:\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002FAntonMu\u002FTrainYourOwnYOLO\ncd TrainYourOwnYOLO\u002F\n```\nCreate Virtual **(Linux\u002FMac)** Environment:\n```bash\npython3 -m venv env\nsource env\u002Fbin\u002Factivate\n```\nMake sure that, from now on, you **run all commands from within your virtual environment**.\n\n#### Setting up Virtual Environment [Windows]\nUse the [Github Desktop GUI](https:\u002F\u002Fdesktop.github.com\u002F) to clone this repo to your local machine. Navigate to the `TrainYourOwnYOLO` project folder and open a power shell window by pressing **Shift + Right Click** and selecting `Open PowerShell window here` in the drop-down menu.\n\nCreate Virtual **(Windows)** Environment:\n\n```powershell\npy -m venv env\n.\\env\\Scripts\\activate\n```\n![PowerShell](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FAntonMu_TrainYourOwnYOLO_readme_8e960a35727d.png)\nMake sure that, from now on, you **run all commands from within your virtual environment**.\n\n#### Install Required Packages [Windows, Mac or Linux]\nInstall required packages (from within your virtual environment) via:\n\n```bash\npip install -r requirements.txt\n```\nIf this fails, you may have to upgrade your pip version first with `pip install pip --upgrade`.\n\n## Quick Start (Inference only)\nTo test the cat face detector on test images located in [`TrainYourOwnYOLO\u002FData\u002FSource_Images\u002FTest_Images`](\u002FData\u002FSource_Images\u002FTest_Images) run the `Minimal_Example.py` script in the root folder with:\n\n```bash\npython Minimal_Example.py\n```\n\nThe outputs are saved in [`TrainYourOwnYOLO\u002FData\u002FSource_Images\u002FTest_Image_Detection_Results`](\u002FData\u002FSource_Images\u002FTest_Image_Detection_Results). This includes:\n - Cat pictures with bounding boxes around faces with confidence scores and\n - [`Detection_Results.csv`](\u002FData\u002FSource_Images\u002FTest_Image_Detection_Results\u002FDetection_Results.csv) file with file names and locations of bounding boxes.\n\n If you want to detect cat faces in your own pictures, replace the cat images in [`Data\u002FSource_Images\u002FTest_Images`](\u002FData\u002FSource_Images\u002FTest_Images) with your own images.\n\n## Full Start (Training and Inference)\n\nTo train your own custom YOLO object detector please follow the instructions detailed in the three numbered subfolders of this repo:\n- [`1_Image_Annotation`](\u002F1_Image_Annotation\u002F),\n- [`2_Training`](\u002F2_Training\u002F) and\n- [`3_Inference`](\u002F3_Inference\u002F).\n \n**To make everything run smoothly it is highly recommended to keep the original folder structure of this repo!**\n\nEach `*.py` script has various command line options that help tweak performance and change things such as input and output directories. All scripts are initialized with good default values that help accomplish all tasks as long as the original folder structure is preserved. To learn more about available command line options of a python script `\u003Cscript_name.py>` run:\n\n```bash\npython \u003Cscript_name.py> -h\n```\n### **NEW:** Weights and Biases\nTrainYourOwnYOLO supports [Weights & Biases](https:\u002F\u002Fwandb.ai\u002Fhome\u002F) to track your experiments online. Sign up at [wandb.ai](https:\u002F\u002Fwandb.ai\u002Fhome) to get an API key and run:\n```bash\nwandb -login \u003CAPI_KEY>\n```\nwhere `\u003CAPI_KEY>` is your Weights & Biases API key. \n\n### Multi-Stream-Multi-Model-Multi-GPU\nIf you want to run multiple streams in parallel, head over to [github.com\u002Fbertelschmitt\u002FmultistreamYOLO](https:\u002F\u002Fgithub.com\u002Fbertelschmitt\u002FmultistreamYOLO). Thanks to @bertelschmitt for putting the work into this.\n\n## License\n\nUnless explicitly stated otherwise at the top of a file, all code is licensed under [CC BY 4.0](https:\u002F\u002Fcreativecommons.org\u002Flicenses\u002Fby\u002F4.0\u002F). This repo makes use of [**ilmonteux\u002Flogohunter**](https:\u002F\u002Fgithub.com\u002Filmonteux\u002Flogohunter) which itself is inspired by [**qqwweee\u002Fkeras-yolo3**](https:\u002F\u002Fgithub.com\u002Fqqwweee\u002Fkeras-yolo3).\n\n## Troubleshooting\n\n0. If you encounter any error, please make sure you follow the instructions **exactly** (word by word). Once you are familiar with the code, you're welcome to modify it as needed but in order to minimize error, I encourage you to not deviate from the instructions above. If you would like to file an issue, please use the provided template and make sure to fill out all fields. \n\n1. If you encounter a `FileNotFoundError`, `Module not found` or similar error, make sure that you did not change the folder structure. Your directory structure **must** look exactly like this: \n    ```text\n    TrainYourOwnYOLO\n    └─── 1_Image_Annotation\n    └─── 2_Training\n    └─── 3_Inference\n    └─── Data\n    └─── Utils\n    ```\n    If you use a different name such as e.g. `TrainYourOwnYOLO-master` you will have to specify the correct paths as command line arguments in every function call.\n\n    Don't use spaces in file or folder names, i.e. instead of `my folder` use `my_folder`.\n\n2. If you are a Linux user and having trouble installing `*.snap` package files try:\n    ```bash\n    snap install --dangerous vott-2.1.0-linux.snap\n    ```\n    See [Snap Tutorial](https:\u002F\u002Ftutorials.ubuntu.com\u002Ftutorial\u002Fadvanced-snap-usage#2) for more information.\n3. If you have a newer version of python on your system, make sure that you create your virtual environment with version 3.7. You can use virtualenv for this: \n    ```\n    pip install virtualenv\n    virtualenv env --python=python3.7\n    ```\n    Then follow the same steps as above. \n## Need more help? File an Issue!\nIf you would like to file an issue, please use the provided issue template and make sure to complete all fields. This makes it easier to reproduce the issue for someone trying to help you. \n\n![Issue](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FAntonMu_TrainYourOwnYOLO_readme_58e78004d73c.gif)\n\nIssues without a completed issue template will be closed and marked with the label \"issue template not completed\". \n\n## Stay Up-to-Date\n\n- ⭐ **star** this repo to get notifications on future improvements and\n- 🍴 **fork** this repo if you like to use it as part of your own project.\n\n![CatVideo](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FAntonMu_TrainYourOwnYOLO_readme_b5ef85595104.gif)\n\n## Licensing \nThis work is licensed under a [Creative Commons Attribution 4.0 International\nLicense][cc-by]. This means that you are free to:\n\n * **Share** — copy and redistribute the material in any medium or format\n * **Adapt** — remix, transform, and build upon the material for any purpose, even commercially.\n\nUnder the following terms:\n\n * **Attribution** \n \n Cite as:\n \n  ```text\n  @misc{TrainYourOwnYOLO,\n    title = {TrainYourOwnYOLO: Building a Custom Object Detector from Scratch},\n    author = {Anton Muehlemann},\n    year = {2019},\n    url = {https:\u002F\u002Fgithub.com\u002FAntonMu\u002FTrainYourOwnYOLO},\n    doi = {10.5281\u002Fzenodo.5112375}\n  }\n  ```\nIf your work doesn't include a citation list, simply link this [github repo](https:\u002F\u002Fgithub.com\u002FAntonMu\u002FTrainYourOwnYOLO)!\n \n[![CC BY 4.0][cc-by-image]][cc-by]\n\n[cc-by]: http:\u002F\u002Fcreativecommons.org\u002Flicenses\u002Fby\u002F4.0\u002F\n[cc-by-image]: https:\u002F\u002Fi.creativecommons.org\u002Fl\u002Fby\u002F4.0\u002F88x31.png\n[cc-by-shield]: https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FLicense-CC%20BY%204.0-lightgrey.svg\n\n","# TrainYourOwnYOLO：从零开始构建自定义目标检测器 [![License: CC BY 4.0](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FLicense-CC%20BY%204.0-lightgrey.svg)](https:\u002F\u002Fcreativecommons.org\u002Flicenses\u002Fby\u002F4.0\u002F) [![DOI](https:\u002F\u002Fzenodo.org\u002Fbadge\u002F197467673.svg)](https:\u002F\u002Fzenodo.org\u002Fbadge\u002Flatestdoi\u002F197467673)\n\n本仓库让你可以使用当前最先进的 [YOLOv3](https:\u002F\u002Fpjreddie.com\u002Fdarknet\u002Fyolo\u002F) 计算机视觉（Computer Vision）算法训练自定义图像检测器。简要介绍可参考这篇 [Medium 文章](https:\u002F\u002Fmedium.com\u002F@muehle\u002Fhow-to-train-your-own-yolov3-detector-from-scratch-224d10e55de2)。本仓库兼容 TensorFlow 2.3 和 Keras 2.4。\n\n### 开始之前：\n\n- 🍴 **Fork** 此仓库，以便将其作为你自己项目的一部分使用。\n- ⭐ **Star** 此仓库，以接收未来改进的通知。\n\n### 流程概览\n\n要构建并测试你的 YOLO 目标检测算法，请按以下步骤操作：\n\n 1. [图像标注（Image Annotation）](\u002F1_Image_Annotation\u002F)\n\t - 安装 Microsoft 的 Visual Object Tagging Tool (VoTT)\n\t - 对图像进行标注\n 2. [训练（Training）](\u002F2_Training\u002F)\n \t- 下载预训练权重（pre-trained weights）\n \t- 在标注好的图像上训练你自己的 YOLO 模型 \n 3. [推理（Inference）](\u002F3_Inference\u002F)\n \t- 在新图像和视频中检测目标\n\n## 仓库结构\n+ [`1_Image_Annotation`](\u002F1_Image_Annotation\u002F)：图像标注的脚本和说明\n+ [`2_Training`](\u002F2_Training\u002F)：训练 YOLOv3 模型的脚本和说明\n+ [`3_Inference`](\u002F3_Inference\u002F)：在新图像和视频上测试已训练 YOLO 模型的脚本和说明\n+ [`Data`](\u002FData\u002F)：输入数据、输出数据、模型权重和结果\n+ [`Utils`](\u002FUtils\u002F)：主脚本所使用的工具脚本\n\n## 快速入门\n\n### Google Colab 教程 \u003Ca href=\"https:\u002F\u002Fcolab.research.google.com\u002Fgithub\u002FAntonMu\u002FTrainYourOwnYOLO\u002Fblob\u002Fmaster\u002FTrainYourOwnYOLO.ipynb\" target=\"_parent\">\u003Cimg src=\"https:\u002F\u002Fcolab.research.google.com\u002Fassets\u002Fcolab-badge.svg\" alt=\"Open In Colab\"\u002F>\u003C\u002Fa>\n使用 Google Colab 可跳过大部分设置步骤，立即开始训练你自己的模型。\n\n### 环境要求\n唯一硬性要求是运行 Python 3.6 或 3.7。要安装 Python 3.7，请访问：\n- [python.org\u002Fdownloads](https:\u002F\u002Fwww.python.org\u002Fdownloads\u002Frelease\u002Fpython-376\u002F) \n\n并按照安装说明操作。注意：本仓库仅在 Python 3.6 和 3.7 上测试过，因此建议使用 `python3.6` 或 `python3.7`。\n\n为了加速训练，建议使用支持 **CUDA** 的 **GPU**。例如，在 [AWS](\u002F2_Training\u002FAWS\u002F) 上可以使用 `p2.xlarge` 实例（配备 Tesla K80 GPU，12GB 显存）。在典型 CPU 上的推理速度约为每秒 2 张图像。如果你想在自己的机器上使用 GPU，请按照 [tensorflow.org\u002Finstall\u002Fgpu](https:\u002F\u002Fwww.tensorflow.org\u002Finstall\u002Fgpu) 的说明安装 CUDA 驱动程序，并确保安装了 [正确版本的 CUDA 和 cuDNN](https:\u002F\u002Fwww.tensorflow.org\u002Finstall\u002Fsource#linux)。\n\n### 安装\n\n#### 设置虚拟环境 [Linux 或 Mac]\n\n使用以下命令克隆本仓库：\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002FAntonMu\u002FTrainYourOwnYOLO\ncd TrainYourOwnYOLO\u002F\n```\n创建虚拟环境 **(Linux\u002FMac)**：\n```bash\npython3 -m venv env\nsource env\u002Fbin\u002Factivate\n```\n请确保从此以后**所有命令都在你的虚拟环境中运行**。\n\n#### 设置虚拟环境 [Windows]\n使用 [GitHub Desktop GUI](https:\u002F\u002Fdesktop.github.com\u002F) 将此仓库克隆到本地。进入 `TrainYourOwnYOLO` 项目文件夹，按住 **Shift + 右键**，在下拉菜单中选择 `在此处打开 PowerShell 窗口`。\n\n创建虚拟环境 **(Windows)**：\n\n```powershell\npy -m venv env\n.\\env\\Scripts\\activate\n```\n![PowerShell](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FAntonMu_TrainYourOwnYOLO_readme_8e960a35727d.png)\n请确保从此以后**所有命令都在你的虚拟环境中运行**。\n\n#### 安装所需包 [Windows、Mac 或 Linux]\n在虚拟环境中通过以下命令安装所需包：\n\n```bash\npip install -r requirements.txt\n```\n如果失败，可能需要先升级 pip：`pip install pip --upgrade`。\n\n## 快速启动（仅推理）\n\n要在 [`TrainYourOwnYOLO\u002FData\u002FSource_Images\u002FTest_Images`](\u002FData\u002FSource_Images\u002FTest_Images) 中的测试图像上测试猫脸检测器，请在根目录运行 `Minimal_Example.py` 脚本：\n\n```bash\npython Minimal_Example.py\n```\n\n输出结果将保存在 [`TrainYourOwnYOLO\u002FData\u002FSource_Images\u002FTest_Image_Detection_Results`](\u002FData\u002FSource_Images\u002FTest_Image_Detection_Results)，包括：\n - 带有置信度分数的猫脸边界框（bounding boxes）的图片，以及\n - 包含文件名和边界框位置的 [`Detection_Results.csv`](\u002FData\u002FSource_Images\u002FTest_Image_Detection_Results\u002FDetection_Results.csv) 文件。\n\n如果你想在自己的图片中检测猫脸，请将 [`Data\u002FSource_Images\u002FTest_Images`](\u002FData\u002FSource_Images\u002FTest_Images) 中的猫图片替换为你自己的图片。\n\n## 完整流程（训练与推理）\n\n要训练你自己的 YOLO 自定义目标检测器，请遵循本仓库三个编号子文件夹中的详细说明：\n- [`1_Image_Annotation`](\u002F1_Image_Annotation\u002F)，\n- [`2_Training`](\u002F2_Training\u002F)，以及\n- [`3_Inference`](\u002F3_Inference\u002F)。\n\n**为确保一切顺利运行，强烈建议保留本仓库的原始文件夹结构！**\n\n每个 `*.py` 脚本都提供了多种命令行选项，可用于调整性能或更改输入输出目录等设置。只要保留原始文件夹结构，所有脚本的默认值都能顺利完成任务。要了解某个 Python 脚本 `\u003Cscript_name.py>` 的可用命令行选项，请运行：\n\n```bash\npython \u003Cscript_name.py> -h\n```\n\n### **新增功能：** Weights and Biases\nTrainYourOwnYOLO 支持使用 [Weights & Biases](https:\u002F\u002Fwandb.ai\u002Fhome\u002F) 在线跟踪你的实验。请在 [wandb.ai](https:\u002F\u002Fwandb.ai\u002Fhome) 注册以获取 API 密钥，并运行：\n```bash\nwandb -login \u003CAPI_KEY>\n```\n其中 `\u003CAPI_KEY>` 是你的 Weights & Biases API 密钥。\n\n### 多流-多模型-多 GPU（Multi-Stream-Multi-Model-Multi-GPU）\n如果你想并行运行多个视频流，请查看 [github.com\u002Fbertelschmitt\u002FmultistreamYOLO](https:\u002F\u002Fgithub.com\u002Fbertelschmitt\u002FmultistreamYOLO)。感谢 @bertelschmitt 的贡献。\n\n## 许可证\n\n除非文件顶部另有明确说明，否则所有代码均采用 [CC BY 4.0](https:\u002F\u002Fcreativecommons.org\u002Flicenses\u002Fby\u002F4.0\u002F) 许可证。本仓库使用了 [**ilmonteux\u002Flogohunter**](https:\u002F\u002Fgithub.com\u002Filmonteux\u002Flogohunter)，而后者又受到 [**qqwweee\u002Fkeras-yolo3**](https:\u002F\u002Fgithub.com\u002Fqqwweee\u002Fkeras-yolo3) 的启发。\n\n## 故障排除（Troubleshooting）\n\n0. 如果你遇到任何错误，请确保你**逐字逐句**地按照说明操作。一旦你熟悉了代码，欢迎根据需要进行修改，但为了尽量减少错误，我建议你不要偏离上述说明。如果你想提交 Issue（问题报告），请使用提供的模板，并确保填写所有字段。\n\n1. 如果你遇到 `FileNotFoundError`、`Module not found` 或类似错误，请确保你没有更改文件夹结构。你的目录结构**必须**完全如下所示：\n    ```text\n    TrainYourOwnYOLO\n    └─── 1_Image_Annotation\n    └─── 2_Training\n    └─── 3_Inference\n    └─── Data\n    └─── Utils\n    ```\n    如果你使用了不同的名称（例如 `TrainYourOwnYOLO-master`），则每次调用函数时都必须通过命令行参数指定正确的路径。\n\n    文件或文件夹名称中不要包含空格，例如应使用 `my_folder` 而非 `my folder`。\n\n2. 如果你是 Linux 用户，并且在安装 `*.snap` 包文件时遇到问题，请尝试：\n    ```bash\n    snap install --dangerous vott-2.1.0-linux.snap\n    ```\n    更多信息请参阅 [Snap 教程](https:\u002F\u002Ftutorials.ubuntu.com\u002Ftutorial\u002Fadvanced-snap-usage#2)。\n\n3. 如果你的系统上安装了更新版本的 Python，请确保使用 3.7 版本创建虚拟环境。你可以使用 virtualenv 来实现：\n    ```\n    pip install virtualenv\n    virtualenv env --python=python3.7\n    ```\n    然后按照上述相同步骤操作。\n\n## 需要更多帮助？提交 Issue！\n如果你想提交 Issue，请使用提供的 Issue 模板，并确保填写所有字段。这有助于其他试图帮助你的人更容易复现问题。\n\n![Issue](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FAntonMu_TrainYourOwnYOLO_readme_58e78004d73c.gif)\n\n未完整填写 Issue 模板的问题将被关闭，并标记为 “issue template not completed”（未完成 Issue 模板）。\n\n## 保持更新\n\n- ⭐ **Star** 此仓库以接收未来改进的通知；\n- 🍴 如果你想将此项目用于自己的项目中，请 **Fork** 此仓库。\n\n![CatVideo](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FAntonMu_TrainYourOwnYOLO_readme_b5ef85595104.gif)\n\n## 许可协议（Licensing）\n本作品采用 [知识共享署名 4.0 国际许可协议（Creative Commons Attribution 4.0 International License）][cc-by] 授权。这意味着你可以自由地：\n\n * **共享（Share）** — 以任何媒介或格式复制和重新分发本素材；\n * **改编（Adapt）** — 对本素材进行混音、转换或在其基础上创作，包括用于商业目的。\n\n但需遵守以下条款：\n\n * **署名（Attribution）**\n\n 引用格式如下：\n\n  ```text\n  @misc{TrainYourOwnYOLO,\n    title = {TrainYourOwnYOLO: Building a Custom Object Detector from Scratch},\n    author = {Anton Muehlemann},\n    year = {2019},\n    url = {https:\u002F\u002Fgithub.com\u002FAntonMu\u002FTrainYourOwnYOLO},\n    doi = {10.5281\u002Fzenodo.5112375}\n  }\n  ```\n如果你的作品中没有引用列表，只需链接到这个 [GitHub 仓库](https:\u002F\u002Fgithub.com\u002FAntonMu\u002FTrainYourOwnYOLO) 即可！\n\n[![CC BY 4.0][cc-by-image]][cc-by]\n\n[cc-by]: http:\u002F\u002Fcreativecommons.org\u002Flicenses\u002Fby\u002F4.0\u002F\n[cc-by-image]: https:\u002F\u002Fi.creativecommons.org\u002Fl\u002Fby\u002F4.0\u002F88x31.png\n[cc-by-shield]: https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FLicense-CC%20BY%204.0-lightgrey.svg","# TrainYourOwnYOLO 快速上手指南\n\n## 环境准备\n\n- **Python 版本**：仅支持 Python 3.6 或 3.7（推荐使用 [清华镜像源](https:\u002F\u002Fmirrors.tuna.tsinghua.edu.cn\u002Fanaconda\u002Farchive\u002F) 安装）\n- **操作系统**：Windows \u002F Linux \u002F macOS 均可\n- **GPU（可选但推荐）**：如需加速训练，建议使用支持 CUDA 的 NVIDIA GPU，并安装对应版本的 [CUDA 和 cuDNN](https:\u002F\u002Fwww.tensorflow.org\u002Finstall\u002Fsource#linux)。国内用户可参考 [清华 CUDA 镜像](https:\u002F\u002Fmirrors.tuna.tsinghua.edu.cn\u002Fhelp\u002Fcuda\u002F)。\n- **依赖库**：TensorFlow 2.3 + Keras 2.4（通过 `requirements.txt` 自动安装）\n\n> 💡 提示：若无本地 GPU，可直接使用 [Google Colab 教程](https:\u002F\u002Fcolab.research.google.com\u002Fgithub\u002FAntonMu\u002FTrainYourOwnYOLO\u002Fblob\u002Fmaster\u002FTrainYourOwnYOLO.ipynb) 免配置运行。\n\n## 安装步骤\n\n### 1. 克隆仓库\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002FAntonMu\u002FTrainYourOwnYOLO\ncd TrainYourOwnYOLO\u002F\n```\n\n### 2. 创建虚拟环境\n\n**Linux \u002F macOS:**\n```bash\npython3 -m venv env\nsource env\u002Fbin\u002Factivate\n```\n\n**Windows (PowerShell):**\n```powershell\npy -m venv env\n.\\env\\Scripts\\activate\n```\n\n> ⚠️ 后续所有命令均需在激活的虚拟环境中执行。\n\n### 3. 安装依赖（推荐使用国内镜像加速）\n\n```bash\npip install -i https:\u002F\u002Fpypi.tuna.tsinghua.edu.cn\u002Fsimple\u002F --upgrade pip\npip install -r requirements.txt\n```\n\n## 基本使用（仅推理）\n\n项目已内置一个猫脸检测模型，可快速测试：\n\n```bash\npython Minimal_Example.py\n```\n\n- **输入图像**：`Data\u002FSource_Images\u002FTest_Images\u002F` 目录下的图片\n- **输出结果**：\n  - 带检测框的图像保存至 `Data\u002FSource_Images\u002FTest_Image_Detection_Results\u002F`\n  - 检测坐标与置信度记录在 `Detection_Results.csv`\n\n> 📌 若想检测自己的图片，只需将你的图像放入 `Data\u002FSource_Images\u002FTest_Images\u002F` 并运行上述命令即可。\n\n> 🔔 注意：请勿修改项目原始目录结构，否则可能导致路径错误。","一家中小型农业无人机公司希望为果园巡检开发一个能自动识别病虫害果实的视觉系统，但团队缺乏深度学习背景，也无现成的检测模型可用。\n\n### 没有 TrainYourOwnYOLO 时\n- 团队需从零搭建 YOLOv3 环境，手动配置 TensorFlow、CUDA、cuDNN 等依赖，耗时数天仍频繁报错。\n- 缺乏清晰的图像标注到模型训练的端到端流程，只能参考零散教程，数据格式转换错误频发。\n- 难以加载预训练权重进行迁移学习，导致在小样本果园图像上训练效果差、收敛慢。\n- 推理阶段需自行编写视频流处理代码，无法快速验证模型在真实飞行视频中的表现。\n- 团队成员因技术门槛高而进展缓慢，项目一度停滞。\n\n### 使用 TrainYourOwnYOLO 后\n- 直接使用其提供的 Google Colab 教程，10 分钟内完成环境配置并开始训练，无需本地 GPU 配置。\n- 按照 `1_Image_Annotation` 到 `3_Inference` 的标准化流程，用 VoTT 标注后一键生成训练数据，格式自动对齐。\n- 内置预训练权重和迁移学习脚本，仅用 200 张果园图像就在 2 小时内训练出准确率超 85% 的定制模型。\n- 利用 `3_Inference` 中的视频检测脚本，直接输入无人机拍摄的 MP4 文件，实时输出带框选结果的视频。\n- 两名工程师一周内完成从数据采集到部署验证的全流程，加速产品原型上线。\n\nTrainYourOwnYOLO 将复杂的 YOLOv3 定制过程封装为清晰、可执行的三步流水线，让非算法团队也能高效构建专属目标检测能力。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FAntonMu_TrainYourOwnYOLO_b5ef8559.gif","AntonMu","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002FAntonMu_6e029490.jpg","University of Oxford,\r\nUC Berkeley","UC Berkeley","Berkeley, USA","antonmuehlemann@gmail.com","antonmuehlemann","antonmuehlemann.de","https:\u002F\u002Fgithub.com\u002FAntonMu",[85,89,93],{"name":86,"color":87,"percentage":88},"Jupyter Notebook","#DA5B0B",98.9,{"name":90,"color":91,"percentage":92},"Python","#3572A5",1.1,{"name":94,"color":95,"percentage":96},"Shell","#89e051",0,674,419,"2026-04-01T19:11:40","NOASSERTION","Linux, macOS, Windows","推荐使用 NVIDIA GPU（如 Tesla K80），显存 12GB，需安装与 TensorFlow 2.3 兼容的 CUDA 和 cuDNN 版本","未说明",{"notes":105,"python":106,"dependencies":107},"必须严格保持项目原始目录结构；训练前需使用 VoTT 工具进行图像标注；支持通过 Weights & Biases 跟踪实验；建议使用虚拟环境管理依赖；Google Colab 可免配置直接运行。","3.6 或 3.7",[108,109,110,111,112,113,114,115,116,117],"tensorflow==2.3","keras==2.4","numpy","pillow","matplotlib","opencv-python","pandas","scipy","h5py","wandb",[13,14],[120,121,122,123,124,125,126,127,128,129,130,131,132,133,134,135,136,137,117,138],"yolov3","yolo","object-detection","python","custom-yolo","deep-learning","deep-learning-tutorial","detector","inference","annotating-images","transfer-learning","google-colab","gpu","tensorflow2","keras","keras-models","trainyourownyolo","tf2","weights-and-biases",4,null,"2026-03-27T02:49:30.150509","2026-04-06T07:11:50.650273",[144,149,154,159,164,168],{"id":145,"question_zh":146,"answer_zh":147,"source_url":148},422,"训练时验证损失（val_loss）为 NaN，如何解决？","可能原因包括学习率过高、输入尺寸过大导致显存不足或配置错误。建议：1）将输入尺寸从 608×608 改为较小的如 224×224；2）确保 cfg 文件中 classes 和 filters 设置正确（filters = (classes + 5) * 3）；3）使用极低学习率（如 1e-10）仅用于调试，实际训练应使用默认或稍高的学习率；4）确认 anchors 已根据数据集重新聚类并正确替换。","https:\u002F\u002Fgithub.com\u002FAntonMu\u002FTrainYourOwnYOLO\u002Fissues\u002F125",{"id":150,"question_zh":151,"answer_zh":152,"source_url":153},423,"运行 Train_YOLO.py 时报错 'No module named keras_yolo3' 怎么办？","该错误通常是因为未正确生成 yolo.h5 模型文件。解决步骤：1）使用 VoTT 标注图像；2）运行 1_image_annotation\u002Fannotate.py 生成标注文件；3）在 2_Training 目录下运行 download_and_convert.py 下载 yolov3.weights 并转换为 yolo.h5；4）确认 yolo.h5 已生成后再运行 train_yolo.py。","https:\u002F\u002Fgithub.com\u002FAntonMu\u002FTrainYourOwnYOLO\u002Fissues\u002F40",{"id":155,"question_zh":156,"answer_zh":157,"source_url":158},424,"使用 TensorFlow-GPU 版本训练时卡在 Epoch 1\u002F51 不动，如何解决？","这是 GPU 内存或兼容性问题。解决方案：1）将 keras 导入替换为 tensorflow.keras；2）在 train_yolo.py 中添加 GPU 内存限制代码：\n```python\nimport tensorflow as tf\nconfig = tf.compat.v1.ConfigProto()\nconfig.gpu_options.per_process_gpu_memory_fraction = 0.7\nconfig.log_device_placement = False\nsess = tf.compat.v1.Session(config=config)\nK.set_session(sess)\n```\n此方法已在 Windows 10 + TensorFlow 2.0 + GTX 1660 上验证有效。","https:\u002F\u002Fgithub.com\u002FAntonMu\u002FTrainYourOwnYOLO\u002Fissues\u002F50",{"id":160,"question_zh":161,"answer_zh":162,"source_url":163},425,"训练结束时出现 'Error occurred when finalizing GeneratorDataset iterator' 警告，是否影响模型？","该警告通常不影响模型可用性，是 TensorFlow 在进程终止时的无害提示。可安全忽略。若训练提前停止（如因 EarlyStopping），建议增加 epochs 数量（如设为 51 或更高）以确保 loss 充分下降，提升精度。","https:\u002F\u002Fgithub.com\u002FAntonMu\u002FTrainYourOwnYOLO\u002Fissues\u002F198",{"id":165,"question_zh":166,"answer_zh":167,"source_url":163},426,"如何确认 TensorFlow 是否正确识别并使用了 GPU？","可在训练脚本中加入以下代码检查 GPU 可用性并启用内存增长：\n```python\nimport tensorflow as tf\nphysical_devices = tf.config.experimental.list_physical_devices('GPU')\nprint(\"physical_devices-------------\", len(physical_devices))\ntf.config.experimental.set_memory_growth(physical_devices[0], True)\n```\n这有助于避免显存分配问题，尤其在多 GPU 或大 batch size 场景下。",{"id":169,"question_zh":170,"answer_zh":171,"source_url":148},427,"训练后模型对未训练类别也给出高置信度检测（如 58%），如何处理？","YOLO 模型可能对未知类别产生误检。建议：1）在推理时提高置信度阈值（如设为 0.7 或更高）；2）确保训练数据覆盖足够多样本，避免过拟合；3）若需严格区分已知\u002F未知类别，可结合背景类训练或使用开放集识别策略。",[173,178],{"id":174,"version":175,"summary_zh":176,"released_at":177},100085,"v0.2.3","Added support for Tensorflow 2.x in compatibility mode.","2021-07-18T18:23:27",{"id":179,"version":180,"summary_zh":181,"released_at":182},100086,"v0.1.15","This is the last revision that works with TensorFlow 1.15 and Keras 2.2.","2020-09-02T06:05:14"]