[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"similar-mgonzs13--yolo_ros":3,"tool-mgonzs13--yolo_ros":64},[4,17,26,40,48,56],{"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},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。",107662,2,"2026-04-03T11:11:01",[13,14,15],{"id":27,"name":28,"github_repo":29,"description_zh":30,"stars":31,"difficulty_score":23,"last_commit_at":32,"category_tags":33,"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",[14,34,35,36,15,37,38,13,39],"数据工具","视频","插件","其他","语言模型","音频",{"id":41,"name":42,"github_repo":43,"description_zh":44,"stars":45,"difficulty_score":10,"last_commit_at":46,"category_tags":47,"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",[15,14,13,38,37],{"id":49,"name":50,"github_repo":51,"description_zh":52,"stars":53,"difficulty_score":10,"last_commit_at":54,"category_tags":55,"status":16},519,"PaddleOCR","PaddlePaddle\u002FPaddleOCR","PaddleOCR 是一款基于百度飞桨框架开发的高性能开源光学字符识别工具包。它的核心能力是将图片、PDF 等文档中的文字提取出来，转换成计算机可读取的结构化数据，让机器真正“看懂”图文内容。\n\n面对海量纸质或电子文档，PaddleOCR 解决了人工录入效率低、数字化成本高的问题。尤其在人工智能领域，它扮演着连接图像与大型语言模型（LLM）的桥梁角色，能将视觉信息直接转化为文本输入，助力智能问答、文档分析等应用场景落地。\n\nPaddleOCR 适合开发者、算法研究人员以及有文档自动化需求的普通用户。其技术优势十分明显：不仅支持全球 100 多种语言的识别，还能在 Windows、Linux、macOS 等多个系统上运行，并灵活适配 CPU、GPU、NPU 等各类硬件。作为一个轻量级且社区活跃的开源项目，PaddleOCR 既能满足快速集成的需求，也能支撑前沿的视觉语言研究，是处理文字识别任务的理想选择。",74913,"2026-04-05T10:44:17",[38,14,13,37],{"id":57,"name":58,"github_repo":59,"description_zh":60,"stars":61,"difficulty_score":23,"last_commit_at":62,"category_tags":63,"status":16},2471,"tesseract","tesseract-ocr\u002Ftesseract","Tesseract 是一款历史悠久且备受推崇的开源光学字符识别（OCR）引擎，最初由惠普实验室开发，后由 Google 维护，目前由全球社区共同贡献。它的核心功能是将图片中的文字转化为可编辑、可搜索的文本数据，有效解决了从扫描件、照片或 PDF 文档中提取文字信息的难题，是数字化归档和信息自动化的重要基础工具。\n\n在技术层面，Tesseract 展现了强大的适应能力。从版本 4 开始，它引入了基于长短期记忆网络（LSTM）的神经网络 OCR 引擎，显著提升了行识别的准确率；同时，为了兼顾旧有需求，它依然支持传统的字符模式识别引擎。Tesseract 原生支持 UTF-8 编码，开箱即用即可识别超过 100 种语言，并兼容 PNG、JPEG、TIFF 等多种常见图像格式。输出方面，它灵活支持纯文本、hOCR、PDF、TSV 等多种格式，方便后续数据处理。\n\nTesseract 主要面向开发者、研究人员以及需要构建文档处理流程的企业用户。由于它本身是一个命令行工具和库（libtesseract），不包含图形用户界面（GUI），因此最适合具备一定编程能力的技术人员集成到自动化脚本或应用程序中",73286,"2026-04-03T01:56:45",[13,14],{"id":65,"github_repo":66,"name":67,"description_en":68,"description_zh":69,"ai_summary_zh":70,"readme_en":71,"readme_zh":72,"quickstart_zh":73,"use_case_zh":74,"hero_image_url":75,"owner_login":76,"owner_name":77,"owner_avatar_url":78,"owner_bio":79,"owner_company":80,"owner_location":81,"owner_email":82,"owner_twitter":83,"owner_website":84,"owner_url":85,"languages":86,"stars":99,"forks":100,"last_commit_at":101,"license":102,"difficulty_score":10,"env_os":103,"env_gpu":104,"env_ram":105,"env_deps":106,"category_tags":115,"github_topics":116,"view_count":23,"oss_zip_url":84,"oss_zip_packed_at":84,"status":16,"created_at":134,"updated_at":135,"faqs":136,"releases":137},2130,"mgonzs13\u002Fyolo_ros","yolo_ros","Ultralytics YOLOv8, YOLOv9, YOLOv10, YOLOv11, YOLOv12 for ROS 2","yolo_ros 是一款专为 ROS 2 机器人系统打造的视觉感知桥梁，它将 Ultralytics 旗下强大的 YOLOv8 至 YOLOv12 系列模型无缝集成到机器人开发环境中。在机器人应用中，如何让机器“看懂”周围世界并实时做出反应往往充满挑战，而 yolo_ros 正是为了解决这一痛点，让开发者无需重复造轮子，即可直接调用最先进的 AI 视觉能力。\n\n这款工具不仅支持基础的物体检测与跟踪，还涵盖了实例分割、人体姿态估计以及定向边界框（OBB）等高级功能。更值得一提的是，它还能结合深度图像实现 3D 物体检测、3D 实例分割及 3D 姿态估计，为机器人提供丰富的空间感知信息。对于从事自动驾驶、服务机器人或工业自动化开发的工程师与研究人员而言，yolo_ros 能显著降低算法部署门槛，加速从原型验证到实际落地的过程。\n\n技术层面，yolo_ros 完美兼容 Humble、Iron 及 Jazzy 等多个主流 ROS 2 发行版，并提供了便捷的 Docker 镜像支持，确保环境配置的一致性与稳定性。无论是希望快速构建智能导航系统的开发者，还是致力于探索复杂场景感知算法的研究人员，都","yolo_ros 是一款专为 ROS 2 机器人系统打造的视觉感知桥梁，它将 Ultralytics 旗下强大的 YOLOv8 至 YOLOv12 系列模型无缝集成到机器人开发环境中。在机器人应用中，如何让机器“看懂”周围世界并实时做出反应往往充满挑战，而 yolo_ros 正是为了解决这一痛点，让开发者无需重复造轮子，即可直接调用最先进的 AI 视觉能力。\n\n这款工具不仅支持基础的物体检测与跟踪，还涵盖了实例分割、人体姿态估计以及定向边界框（OBB）等高级功能。更值得一提的是，它还能结合深度图像实现 3D 物体检测、3D 实例分割及 3D 姿态估计，为机器人提供丰富的空间感知信息。对于从事自动驾驶、服务机器人或工业自动化开发的工程师与研究人员而言，yolo_ros 能显著降低算法部署门槛，加速从原型验证到实际落地的过程。\n\n技术层面，yolo_ros 完美兼容 Humble、Iron 及 Jazzy 等多个主流 ROS 2 发行版，并提供了便捷的 Docker 镜像支持，确保环境配置的一致性与稳定性。无论是希望快速构建智能导航系统的开发者，还是致力于探索复杂场景感知算法的研究人员，都能通过 yolo_ros 轻松获取高精度的实时视觉分析能力，让机器人真正具备敏锐的“眼睛”。","# yolo_ros\n\nROS 2 wrap for YOLO models from [Ultralytics](https:\u002F\u002Fgithub.com\u002Fultralytics\u002Fultralytics) to perform object detection and tracking, instance segmentation, human pose estimation and Oriented Bounding Box (OBB). There are also 3D versions of object detection, including instance segmentation, and human pose estimation based on depth images.\n\n\u003Cdiv align=\"center\">\n\n[![License: GPL](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FGitHub-GPL--3.0-informational)](https:\u002F\u002Fopensource.org\u002Flicense\u002Fgpl-3-0) [![GitHub release](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Frelease\u002Fmgonzs13\u002Fyolo_ros.svg)](https:\u002F\u002Fgithub.com\u002Fmgonzs13\u002Fyolo_ros\u002Freleases) [![Code Size](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Flanguages\u002Fcode-size\u002Fmgonzs13\u002Fyolo_ros.svg?branch=main)](https:\u002F\u002Fgithub.com\u002Fmgonzs13\u002Fyolo_ros?branch=main) [![Dependencies](https:\u002F\u002Fimg.shields.io\u002Flibrariesio\u002Fgithub\u002Fmgonzs13\u002Fyolo_ros?branch=main)](https:\u002F\u002Flibraries.io\u002Fgithub\u002Fmgonzs13\u002Fyolo_ros?branch=main) [![Last Commit](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Flast-commit\u002Fmgonzs13\u002Fyolo_ros.svg)](https:\u002F\u002Fgithub.com\u002Fmgonzs13\u002Fyolo_ros\u002Fcommits\u002Fmain) [![GitHub issues](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fissues\u002Fmgonzs13\u002Fyolo_ros)](https:\u002F\u002Fgithub.com\u002Fmgonzs13\u002Fyolo_ros\u002Fissues) [![GitHub pull requests](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fissues-pr\u002Fmgonzs13\u002Fyolo_ros)](https:\u002F\u002Fgithub.com\u002Fmgonzs13\u002Fyolo_ros\u002Fpulls) [![Contributors](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fcontributors\u002Fmgonzs13\u002Fyolo_ros.svg)](https:\u002F\u002Fgithub.com\u002Fmgonzs13\u002Fyolo_ros\u002Fgraphs\u002Fcontributors) [![Python Formatter Check](https:\u002F\u002Fgithub.com\u002Fmgonzs13\u002Fyolo_ros\u002Factions\u002Fworkflows\u002Fpython-formatter.yml\u002Fbadge.svg?branch=main)](https:\u002F\u002Fgithub.com\u002Fmgonzs13\u002Fyolo_ros\u002Factions\u002Fworkflows\u002Fpython-formatter.yml?branch=main) [![Doxygen Deployment](https:\u002F\u002Fgithub.com\u002Fmgonzs13\u002Fyolo_ros\u002Factions\u002Fworkflows\u002Fdoxygen-deployment.yml\u002Fbadge.svg)](https:\u002F\u002Fmgonzs13.github.io\u002Fyolo_ros\u002Flatest)\n\n| ROS 2 Distro |                          Branch                          |                                                                                                      Build status                                                                                                      |                                                               Docker Image                                                                |\n| :----------: | :------------------------------------------------------: | :--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------: | :---------------------------------------------------------------------------------------------------------------------------------------: |\n|  **Humble**  | [`main`](https:\u002F\u002Fgithub.com\u002Fmgonzs13\u002Fyolo_ros\u002Ftree\u002Fmain) |  [![Humble Build](https:\u002F\u002Fgithub.com\u002Fmgonzs13\u002Fyolo_ros\u002Factions\u002Fworkflows\u002Fhumble-docker-build.yml\u002Fbadge.svg?branch=main)](https:\u002F\u002Fgithub.com\u002Fmgonzs13\u002Fyolo_ros\u002Factions\u002Fworkflows\u002Fhumble-docker-build.yml?branch=main)   |  [![Docker Image](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FDocker%20Image%20-humble-blue)](https:\u002F\u002Fhub.docker.com\u002Fr\u002Fmgons\u002Fyolo_ros\u002Ftags?name=humble)  |\n|   **Iron**   | [`main`](https:\u002F\u002Fgithub.com\u002Fmgonzs13\u002Fyolo_ros\u002Ftree\u002Fmain) |     [![Iron Build](https:\u002F\u002Fgithub.com\u002Fmgonzs13\u002Fyolo_ros\u002Factions\u002Fworkflows\u002Firon-docker-build.yml\u002Fbadge.svg?branch=main)](https:\u002F\u002Fgithub.com\u002Fmgonzs13\u002Fyolo_ros\u002Factions\u002Fworkflows\u002Firon-docker-build.yml?branch=main)      |    [![Docker Image](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FDocker%20Image%20-iron-blue)](https:\u002F\u002Fhub.docker.com\u002Fr\u002Fmgons\u002Fyolo_ros\u002Ftags?name=iron)    |\n|  **Jazzy**   | [`main`](https:\u002F\u002Fgithub.com\u002Fmgonzs13\u002Fyolo_ros\u002Ftree\u002Fmain) |    [![Jazzy Build](https:\u002F\u002Fgithub.com\u002Fmgonzs13\u002Fyolo_ros\u002Factions\u002Fworkflows\u002Fjazzy-docker-build.yml\u002Fbadge.svg?branch=main)](https:\u002F\u002Fgithub.com\u002Fmgonzs13\u002Fyolo_ros\u002Factions\u002Fworkflows\u002Fjazzy-docker-build.yml?branch=main)    |   [![Docker Image](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FDocker%20Image%20-jazzy-blue)](https:\u002F\u002Fhub.docker.com\u002Fr\u002Fmgons\u002Fyolo_ros\u002Ftags?name=jazzy)   |\n|  **Kilted**  | [`main`](https:\u002F\u002Fgithub.com\u002Fmgonzs13\u002Fyolo_ros\u002Ftree\u002Fmain) |  [![Kilted Build](https:\u002F\u002Fgithub.com\u002Fmgonzs13\u002Fyolo_ros\u002Factions\u002Fworkflows\u002Fkilted-docker-build.yml\u002Fbadge.svg?branch=main)](https:\u002F\u002Fgithub.com\u002Fmgonzs13\u002Fyolo_ros\u002Factions\u002Fworkflows\u002Fkilted-docker-build.yml?branch=main)   |  [![Docker Image](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FDocker%20Image%20-kilted-blue)](https:\u002F\u002Fhub.docker.com\u002Fr\u002Fmgons\u002Fyolo_ros\u002Ftags?name=kilted)  |\n| **Rolling**  | [`main`](https:\u002F\u002Fgithub.com\u002Fmgonzs13\u002Fyolo_ros\u002Ftree\u002Fmain) | [![Rolling Build](https:\u002F\u002Fgithub.com\u002Fmgonzs13\u002Fyolo_ros\u002Factions\u002Fworkflows\u002Frolling-docker-build.yml\u002Fbadge.svg?branch=main)](https:\u002F\u002Fgithub.com\u002Fmgonzs13\u002Fyolo_ros\u002Factions\u002Fworkflows\u002Frolling-docker-build.yml?branch=main) | [![Docker Image](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FDocker%20Image%20-rolling-blue)](https:\u002F\u002Fhub.docker.com\u002Fr\u002Fmgons\u002Fyolo_ros\u002Ftags?name=rolling) |\n\n\u003C\u002Fdiv>\n\n## Table of Contents\n\n1. [Installation](#installation)\n2. [Docker](#docker)\n3. [Models](#models)\n4. [Usage](#usage)\n5. [Lifecycle Nodes](#lifecycle-nodes)\n6. [Demos](#demos)\n\n## Installation\n\n```shell\ncd ~\u002Fros2_ws\u002Fsrc\ngit clone https:\u002F\u002Fgithub.com\u002Fmgonzs13\u002Fyolo_ros.git\npip3 install -r yolo_ros\u002Frequirements.txt\ncd ~\u002Fros2_ws\nrosdep install --from-paths src --ignore-src -r -y\ncolcon build\nsource ~\u002Fros2_ws\u002Finstall\u002Fsetup.bash\"\n```\n\n## Docker\n\nBuild the yolo_ros docker.\n\n```shell\ndocker build -t yolo_ros .\n```\n\nRun the docker container. If you want to use CUDA, you have to install the [NVIDIA Container Tollkit](https:\u002F\u002Fdocs.nvidia.com\u002Fdatacenter\u002Fcloud-native\u002Fcontainer-toolkit\u002Flatest\u002Finstall-guide.html) and add `--gpus all`.\n\n```shell\ndocker run -it --rm --gpus all yolo_ros\n```\n\n## Models\n\nThe compatible models for yolo_ros are the following:\n\n- [YOLOv3](https:\u002F\u002Fdocs.ultralytics.com\u002Fmodels\u002Fyolov3\u002F)\n- [YOLOv4](https:\u002F\u002Fdocs.ultralytics.com\u002Fmodels\u002Fyolov4\u002F)\n- [YOLOv5](https:\u002F\u002Fdocs.ultralytics.com\u002Fmodels\u002Fyolov5\u002F)\n- [YOLOv6](https:\u002F\u002Fdocs.ultralytics.com\u002Fmodels\u002Fyolov6\u002F)\n- [YOLOv7](https:\u002F\u002Fdocs.ultralytics.com\u002Fmodels\u002Fyolov7\u002F)\n- [YOLOv8](https:\u002F\u002Fdocs.ultralytics.com\u002Fmodels\u002Fyolov8\u002F)\n- [YOLOv9](https:\u002F\u002Fdocs.ultralytics.com\u002Fmodels\u002Fyolov9\u002F)\n- [YOLOv10](https:\u002F\u002Fdocs.ultralytics.com\u002Fmodels\u002Fyolov10\u002F)\n- [YOLOv11](https:\u002F\u002Fdocs.ultralytics.com\u002Fmodels\u002Fyolo11\u002F)\n- [YOLOv12](https:\u002F\u002Fdocs.ultralytics.com\u002Fmodels\u002Fyolo12\u002F)\n- [YOLO-World](https:\u002F\u002Fdocs.ultralytics.com\u002Fmodels\u002Fyolo-world\u002F)\n- [YOLOE](https:\u002F\u002Fdocs.ultralytics.com\u002Fmodels\u002Fyoloe\u002F)\n\n## Usage\n\n\u003Cdetails>\n\u003Csummary>Click to expand\u003C\u002Fsummary>\n\n### YOLOv5\n\n```shell\nros2 launch yolo_bringup yolov5.launch.py\n```\n\n### YOLOv8\n\n```shell\nros2 launch yolo_bringup yolov8.launch.py\n```\n\n### YOLOv9\n\n```shell\nros2 launch yolo_bringup yolov9.launch.py\n```\n\n### YOLOv10\n\n```shell\nros2 launch yolo_bringup yolov10.launch.py\n```\n\n### YOLOv11\n\n```shell\nros2 launch yolo_bringup yolov11.launch.py\n```\n\n### YOLOv12\n\n```shell\nros2 launch yolo_bringup yolov12.launch.py\n```\n\n### YOLO-World\n\n```shell\nros2 launch yolo_bringup yolo-world.launch.py\n```\n\n### YOLOE\n\n```shell\nros2 launch yolo_bringup yoloe.launch.py\n```\n\n\u003C\u002Fdetails>\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmgonzs13_yolo_ros_readme_0309aefe81d8.png\" width=\"100%\" \u002F>\n\u003C\u002Fp>\n\n### Topics\n\n- **\u002Fyolo\u002Fdetections**: Objects detected by YOLO using the RGB images. Each object contains a bounding box and a class name. It may also include a mark or a list of keypoints.\n- **\u002Fyolo\u002Ftracking**: Objects detected and tracked from YOLO results. Each object is assigned a tracking ID.\n- **\u002Fyolo\u002Fdetections_3d**: 3D objects detected. YOLO results are used to crop the depth images to create the 3D bounding boxes and 3D keypoints.\n- **\u002Fyolo\u002Fdebug_image**: Debug images showing the detected and tracked objects. They can be visualized with rviz2.\n\n### Parameters\n\nThese are the parameters from the [yolo.launch.py](.\u002Fyolo_bringup\u002Flaunch\u002Fyolo.launch.py), used to launch all models. Check out the [Ultralytics page](https:\u002F\u002Fdocs.ultralytics.com\u002Fmodes\u002Fpredict\u002F#inference-arguments) for more details.\n\n- **model_type**: Ultralytics model type (default: YOLO)\n- **model**: YOLO model (default: yolov8m.pt)\n- **tracker**: Tracker file (default: bytetrack.yaml)\n- **device**: GPU\u002FCUDA (default: cuda:0)\n- **fuse_model**: Whether to fuse the YOLO model for inference optimization (default: False)\n- **yolo_encoding**: Encoding to convert input image before using YOLO (default: bgr8)\n- **enable**: Whether to start YOLO enabled (default: True)\n- **threshold**: Detection threshold (default: 0.5)\n- **iou**: Intersection Over Union (IoU) threshold for Non-Maximum Suppression (NMS) (default: 0.7)\n- **imgsz_height**: Image height for inference (default: 480)\n- **imgsz_width**: Image width for inference (default: 640)\n- **half**: Whether to enable half-precision (FP16) inference speeding up model inference with minimal impact on accuracy (default: False)\n- **max_det**: Maximum number of detections allowed per image (default: 300)\n- **augment**: Whether to enable test-time augmentation (TTA) for predictions improving detection robustness at the cost of speed (default: False)\n- **agnostic_nms**: Whether to enable class-agnostic Non-Maximum Suppression (NMS) merging overlapping boxes of different classes (default: False)\n- **retina_masks**: Whether to use high-resolution segmentation masks if available in the model, enhancing mask quality for segmentation (default: False)\n- **input_image_topic**: Camera topic of RGB images (default: \u002Fcamera\u002Frgb\u002Fimage_raw)\n- **image_reliability**: Reliability for the image topic: 0=system default, 1=Reliable, 2=Best Effort (default: 1)\n- **input_depth_topic**: Camera topic of depth images (default: \u002Fcamera\u002Fdepth\u002Fimage_raw)\n- **depth_image_reliability**: Reliability for the depth image topic: 0=system default, 1=Reliable, 2=Best Effort (default: 1)\n- **input_depth_info_topic**: Camera topic for info data (default: \u002Fcamera\u002Fdepth\u002Fcamera_info)\n- **depth_info_reliability**: Reliability for the depth info topic: 0=system default, 1=Reliable, 2=Best Effort (default: 1)\n- **target_frame**: frame to transform the 3D boxes (default: base_link)\n- **depth_image_units_divisor**: Divisor to convert the depth image into meters. Depends on the camera you are using (default: 1000)\n- **use_tracking**: Whether to activate tracking after detection (default: True)\n- **use_3d**: Whether to activate 3D detections (default: False)\n- **use_debug**: Whether to activate debug node (default: True)\n\n## Lifecycle Nodes\n\nPrevious updates add Lifecycle Nodes support to all the nodes available in the package.\nThis implementation tries to reduce the workload in the unconfigured and inactive states by only loading the models and activating the subscriber on the active state.\n\nThese are some resource comparisons using the default yolov8m.pt model on a 30fps video stream.\n\n| State    | CPU Usage (i7 12th Gen) | VRAM Usage | Bandwidth Usage |\n| -------- | ----------------------- | ---------- | --------------- |\n| Active   | 40-50% in one core      | 628 MB     | Up to 200 Mbps  |\n| Inactive | ~5-7% in one core       | 338 MB     | 0-20 Kbps       |\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmgonzs13_yolo_ros_readme_f75e422ced5a.png\" width=\"100%\" \u002F>\n\u003C\u002Fp>\n\n## Demos\n\n## Object Detection\n\nThis is the standard behavior of yolo_ros which includes object tracking.\n\n```shell\nros2 launch yolo_bringup yolo.launch.py\n```\n\n[![](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmgonzs13_yolo_ros_readme_967a3dbe83d2.png)](https:\u002F\u002Fdrive.google.com\u002Ffile\u002Fd\u002F1gTQt6soSIq1g2QmK7locHDiZ-8MqVl2w\u002Fview?usp=sharing)\n\n## Instance Segmentation\n\nInstance masks are the borders of the detected objects, not all the pixels inside the masks.\n\n```shell\nros2 launch yolo_bringup yolo.launch.py model:=yolov8m-seg.pt\n```\n\n[![](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmgonzs13_yolo_ros_readme_02f9abf4bf5d.png)](https:\u002F\u002Fdrive.google.com\u002Ffile\u002Fd\u002F1dwArjDLSNkuOGIB0nSzZR6ABIOCJhAFq\u002Fview?usp=sharing)\n\n## Human Pose\n\nVisible persons are detected along with their skeleton keypoints.\n\n```shell\nros2 launch yolo_bringup yolo.launch.py model:=yolov8m-pose.pt\n```\n\n[![](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmgonzs13_yolo_ros_readme_8a487dd3501e.png)](https:\u002F\u002Fdrive.google.com\u002Ffile\u002Fd\u002F1pRy9lLSXiFEVFpcbesMCzmTMEoUXGWgr\u002Fview?usp=sharing)\n\n## 3D Object Detection\n\nThe 3D bounding boxes are calculated by filtering the depth image data from an RGB-D camera using the 2D bounding box. Only objects with a 3D bounding box are visualized in the 2D image.\n\n```shell\nros2 launch yolo_bringup yolo.launch.py use_3d:=True\n```\n\n[![](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmgonzs13_yolo_ros_readme_4e1d9c06578e.png)](https:\u002F\u002Fdrive.google.com\u002Ffile\u002Fd\u002F1ZcN_u9RB9_JKq37mdtpzXx3b44tlU-pr\u002Fview?usp=sharing)\n\n## 3D Object Detection (Using Instance Segmentation Masks)\n\nIn this, the depth image data is filtered using the max and min values obtained from the instance masks. Only objects with a 3D bounding box are visualized in the 2D image.\n\n```shell\nros2 launch yolo_bringup yolo.launch.py model:=yolov8m-seg.pt use_3d:=True\n```\n\n[![](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmgonzs13_yolo_ros_readme_e1ec8b052fc6.png)](https:\u002F\u002Fdrive.google.com\u002Ffile\u002Fd\u002F1wVZgi5GLkAYxv3GmTxX5z-vB8RQdwqLP\u002Fview?usp=sharing)\n\n## 3D Human Pose\n\nEach keypoint is projected in the depth image and visualized using purple spheres. Only objects with a 3D bounding box are visualized in the 2D image.\n\n```shell\nros2 launch yolo_bringup yolo.launch.py model:=yolov8m-pose.pt use_3d:=True\n```\n\n[![](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmgonzs13_yolo_ros_readme_61fea9f07e43.png)](https:\u002F\u002Fdrive.google.com\u002Ffile\u002Fd\u002F1j4VjCAsOCx_mtM2KFPOLkpJogM0t227r\u002Fview?usp=sharing)\n","# yolo_ros\n\n针对 [Ultralytics](https:\u002F\u002Fgithub.com\u002Fultralytics\u002Fultralytics) 提供的 YOLO 模型的 ROS 2 封装，用于执行目标检测与跟踪、实例分割、人体姿态估计以及定向边界框（OBB）任务。此外，还支持基于深度图像的目标检测、实例分割和人体姿态估计的 3D 版本。\n\n\u003Cdiv align=\"center\">\n\n[![License: GPL](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FGitHub-GPL--3.0-informational)](https:\u002F\u002Fopensource.org\u002Flicense\u002Fgpl-3-0) [![GitHub release](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Frelease\u002Fmgonzs13\u002Fyolo_ros.svg)](https:\u002F\u002Fgithub.com\u002Fmgonzs13\u002Fyolo_ros\u002Freleases) [![Code Size](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Flanguages\u002Fcode-size\u002Fmgonzs13\u002Fyolo_ros.svg?branch=main)](https:\u002F\u002Fgithub.com\u002Fmgonzs13\u002Fyolo_ros?branch=main) [![Dependencies](https:\u002F\u002Fimg.shields.io\u002Flibrariesio\u002Fgithub\u002Fmgonzs13\u002Fyolo_ros?branch=main)](https:\u002F\u002Flibraries.io\u002Fgithub\u002Fmgonzs13\u002Fyolo_ros?branch=main) [![Last Commit](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Flast-commit\u002Fmgonzs13\u002Fyolo_ros.svg)](https:\u002F\u002Fgithub.com\u002Fmgonzs13\u002Fyolo_ros\u002Fcommits\u002Fmain) [![GitHub issues](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fissues\u002Fmgonzs13\u002Fyolo_ros)](https:\u002F\u002Fgithub.com\u002Fmgonzs13\u002Fyolo_ros\u002Fissues) [![GitHub pull requests](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fissues-pr\u002Fmgonzs13\u002Fyolo_ros)](https:\u002F\u002Fgithub.com\u002Fmgonzs13\u002Fyolo_ros\u002Fpulls) [![Contributors](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fcontributors\u002Fmgonzs13\u002Fyolo_ros.svg)](https:\u002F\u002Fgithub.com\u002Fmgonzs13\u002Fyolo_ros\u002Fgraphs\u002Fcontributors) [![Python Formatter Check](https:\u002F\u002Fgithub.com\u002Fmgonzs13\u002Fyolo_ros\u002Factions\u002Fworkflows\u002Fpython-formatter.yml\u002Fbadge.svg?branch=main)](https:\u002F\u002Fgithub.com\u002Fmgonzs13\u002Fyolo_ros\u002Factions\u002Fworkflows\u002Fpython-formatter.yml?branch=main) [![Doxygen Deployment](https:\u002F\u002Fgithub.com\u002Fmgonzs13\u002Fyolo_ros\u002Factions\u002Fworkflows\u002Fdoxygen-deployment.yml\u002Fbadge.svg)](https:\u002F\u002Fmgonzs13.github.io\u002Fyolo_ros\u002Flatest)\n\n| ROS 2 发行版 |                          分支                          |                                                                                                      构建状态                                                                                                      |                                                               Docker 镜像                                                                |\n| :----------: | :------------------------------------------------------: | :--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------: | :---------------------------------------------------------------------------------------------------------------------------------------: |\n|  **Humble**  | [`main`](https:\u002F\u002Fgithub.com\u002Fmgonzs13\u002Fyolo_ros\u002Ftree\u002Fmain) |  [![Humble 构建](https:\u002F\u002Fgithub.com\u002Fmgonzs13\u002Fyolo_ros\u002Factions\u002Fworkflows\u002Fhumble-docker-build.yml\u002Fbadge.svg?branch=main)](https:\u002F\u002Fgithub.com\u002Fmgonzs13\u002Fyolo_ros\u002Factions\u002Fworkflows\u002Fhumble-docker-build.yml?branch=main)   |  [![Docker 镜像](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FDocker%20Image%20-humble-blue)](https:\u002F\u002Fhub.docker.com\u002Fr\u002Fmgons\u002Fyolo_ros\u002Ftags?name=humble)  |\n|   **Iron**   | [`main`](https:\u002F\u002Fgithub.com\u002Fmgonzs13\u002Fyolo_ros\u002Ftree\u002Fmain) |     [![Iron 构建](https:\u002F\u002Fgithub.com\u002Fmgonzs13\u002Fyolo_ros\u002Factions\u002Fworkflows\u002Firon-docker-build.yml\u002Fbadge.svg?branch=main)](https:\u002F\u002Fgithub.com\u002Fmgonzs13\u002Fyolo_ros\u002Factions\u002Fworkflows\u002Firon-docker-build.yml?branch=main)      |    [![Docker 镜像](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FDocker%20Image%20-iron-blue)](https:\u002F\u002Fhub.docker.com\u002Fr\u002Fmgons\u002Fyolo_ros\u002Ftags?name=iron)    |\n|  **Jazzy**   | [`main`](https:\u002F\u002Fgithub.com\u002Fmgonzs13\u002Fyolo_ros\u002Ftree\u002Fmain) |    [![Jazzy 构建](https:\u002F\u002Fgithub.com\u002Fmgonzs13\u002Fyolo_ros\u002Factions\u002Fworkflows\u002Fjazzy-docker-build.yml\u002Fbadge.svg?branch=main)](https:\u002F\u002Fgithub.com\u002Fmgonzs13\u002Fyolo_ros\u002Factions\u002Fworkflows\u002Fjazzy-docker-build.yml?branch=main)    |   [![Docker 镜像](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FDocker%20Image%20-jazzy-blue)](https:\u002F\u002Fhub.docker.com\u002Fr\u002Fmgons\u002Fyolo_ros\u002Ftags?name=jazzy)   |\n|  **Kilted**  | [`main`](https:\u002F\u002Fgithub.com\u002Fmgonzs13\u002Fyolo_ros\u002Ftree\u002Fmain) |  [![Kilted 构建](https:\u002F\u002Fgithub.com\u002Fmgonzs13\u002Fyolo_ros\u002Factions\u002Fworkflows\u002Fkilted-docker-build.yml\u002Fbadge.svg?branch=main)](https:\u002F\u002Fgithub.com\u002Fmgonzs13\u002Fyolo_ros\u002Factions\u002Fworkflows\u002Fkilted-docker-build.yml?branch=main)   |  [![Docker 镜像](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FDocker%20Image%20-kilted-blue)](https:\u002F\u002Fhub.docker.com\u002Fr\u002Fmgons\u002Fyolo_ros\u002Ftags?name=kilted)  |\n| **Rolling**  | [`main`](https:\u002F\u002Fgithub.com\u002Fmgonzs13\u002Fyolo_ros\u002Ftree\u002Fmain) | [![Rolling 构建](https:\u002F\u002Fgithub.com\u002Fmgonzs13\u002Fyolo_ros\u002Factions\u002Fworkflows\u002Frolling-docker-build.yml\u002Fbadge.svg?branch=main)](https:\u002F\u002Fgithub.com\u002Fmgonzs13\u002Fyolo_ros\u002Factions\u002Fworkflows\u002Frolling-docker-build.yml?branch=main) | [![Docker 镜像](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FDocker%20Image%20-rolling-blue)](https:\u002F\u002Fhub.docker.com\u002Fr\u002Fmgons\u002Fyolo_ros\u002Ftags?name=rolling) |\n\n\u003C\u002Fdiv>\n\n## 目录\n\n1. [安装](#installation)\n2. [Docker](#docker)\n3. [模型](#models)\n4. [使用](#usage)\n5. [生命周期节点](#lifecycle-nodes)\n6. [演示](#demos)\n\n## 安装\n\n```shell\ncd ~\u002Fros2_ws\u002Fsrc\ngit clone https:\u002F\u002Fgithub.com\u002Fmgonzs13\u002Fyolo_ros.git\npip3 install -r yolo_ros\u002Frequirements.txt\ncd ~\u002Fros2_ws\nrosdep install --from-paths src --ignore-src -r -y\ncolcon build\nsource ~\u002Fros2_ws\u002Finstall\u002Fsetup.bash\"\n```\n\n## Docker\n\n构建 yolo_ros 的 Docker 镜像。\n\n```shell\ndocker build -t yolo_ros .\n```\n\n运行 Docker 容器。如果需要使用 CUDA，必须先安装 [NVIDIA Container Toolkit](https:\u002F\u002Fdocs.nvidia.com\u002Fdatacenter\u002Fcloud-native\u002Fcontainer-toolkit\u002Flatest\u002Finstall-guide.html)，并在运行命令中添加 `--gpus all`。\n\n```shell\ndocker run -it --rm --gpus all yolo_ros\n```\n\n## 模型\n\nyolo_ros 支持以下模型：\n\n- [YOLOv3](https:\u002F\u002Fdocs.ultralytics.com\u002Fmodels\u002Fyolov3\u002F)\n- [YOLOv4](https:\u002F\u002Fdocs.ultralytics.com\u002Fmodels\u002Fyolov4\u002F)\n- [YOLOv5](https:\u002F\u002Fdocs.ultralytics.com\u002Fmodels\u002Fyolov5\u002F)\n- [YOLOv6](https:\u002F\u002Fdocs.ultralytics.com\u002Fmodels\u002Fyolov6\u002F)\n- [YOLOv7](https:\u002F\u002Fdocs.ultralytics.com\u002Fmodels\u002Fyolov7\u002F)\n- [YOLOv8](https:\u002F\u002Fdocs.ultralytics.com\u002Fmodels\u002Fyolov8\u002F)\n- [YOLOv9](https:\u002F\u002Fdocs.ultralytics.com\u002Fmodels\u002Fyolov9\u002F)\n- [YOLOv10](https:\u002F\u002Fdocs.ultralytics.com\u002Fmodels\u002Fyolov10\u002F)\n- [YOLOv11](https:\u002F\u002Fdocs.ultralytics.com\u002Fmodels\u002Fyolo11\u002F)\n- [YOLOv12](https:\u002F\u002Fdocs.ultralytics.com\u002Fmodels\u002Fyolo12\u002F)\n- [YOLO-World](https:\u002F\u002Fdocs.ultralytics.com\u002Fmodels\u002Fyolo-world\u002F)\n- [YOLOE](https:\u002F\u002Fdocs.ultralytics.com\u002Fmodels\u002Fyoloe\u002F)\n\n## 使用\n\n\u003Cdetails>\n\u003Csummary>点击展开\u003C\u002Fsummary>\n\n### YOLOv5\n\n```shell\nros2 launch yolo_bringup yolov5.launch.py\n```\n\n### YOLOv8\n\n```shell\nros2 launch yolo_bringup yolov8.launch.py\n```\n\n### YOLOv9\n\n```shell\nros2 launch yolo_bringup yolov9.launch.py\n```\n\n### YOLOv10\n\n```shell\nros2 launch yolo_bringup yolov10.launch.py\n```\n\n### YOLOv11\n\n```shell\nros2 launch yolo_bringup yolov11.launch.py\n```\n\n### YOLOv12\n\n```shell\nros2 launch yolo_bringup yolov12.launch.py\n```\n\n### YOLO-World\n\n```shell\nros2 launch yolo_bringup yolo-world.launch.py\n```\n\n### YOLOE\n\n```shell\nros2 launch yolo_bringup yoloe.launch.py\n```\n\n\u003C\u002Fdetails>\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmgonzs13_yolo_ros_readme_0309aefe81d8.png\" width=\"100%\" \u002F>\n\u003C\u002Fp>\n\n### Topics\n\n- **\u002Fyolo\u002Fdetections**: 使用 RGB 图像由 YOLO 检测到的物体。每个物体包含一个边界框和类别名称，也可能包括标记或关键点列表。\n- **\u002Fyolo\u002Ftracking**: 基于 YOLO 结果检测并跟踪的物体。每个物体都会被分配一个跟踪 ID。\n- **\u002Fyolo\u002Fdetections_3d**: 检测到的 3D 物体。利用 YOLO 的结果裁剪深度图像，生成 3D 边界框和 3D 关键点。\n- **\u002Fyolo\u002Fdebug_image**: 显示检测和跟踪物体的调试图像。这些图像可以通过 rviz2 进行可视化。\n\n### Parameters\n\n以下是来自 [yolo.launch.py](.\u002Fyolo_bringup\u002Flaunch\u002Fyolo.launch.py) 的参数，用于启动所有模型。更多详细信息请参阅 [Ultralytics 官网](https:\u002F\u002Fdocs.ultralytics.com\u002Fmodes\u002Fpredict\u002F#inference-arguments)。\n\n- **model_type**: Ultralytics 模型类型（默认：YOLO）\n- **model**: YOLO 模型（默认：yolov8m.pt）\n- **tracker**: 跟踪器文件（默认：bytetrack.yaml）\n- **device**: GPU\u002FCUDA（默认：cuda:0）\n- **fuse_model**: 是否融合 YOLO 模型以优化推理性能（默认：False）\n- **yolo_encoding**: 在使用 YOLO 之前对输入图像进行编码的格式（默认：bgr8）\n- **enable**: 是否启用 YOLO（默认：True）\n- **threshold**: 检测阈值（默认：0.5）\n- **iou**: 非极大值抑制 (NMS) 的交并比 (IoU) 阈值（默认：0.7）\n- **imgsz_height**: 推理时使用的图像高度（默认：480）\n- **imgsz_width**: 推理时使用的图像宽度（默认：640）\n- **half**: 是否启用半精度 (FP16) 推理，以在对精度影响最小的情况下加速模型推理（默认：False）\n- **max_det**: 每张图像允许的最大检测数量（默认：300）\n- **augment**: 是否启用测试时增强 (TTA)，以提高检测鲁棒性，但会牺牲部分速度（默认：False）\n- **agnostic_nms**: 是否启用类别无关的非极大值抑制 (NMS)，将不同类别的重叠框合并（默认：False）\n- **retina_masks**: 是否使用高分辨率分割掩码（如果模型中可用），以提升分割掩码的质量（默认：False）\n- **input_image_topic**: RGB 图像的相机话题（默认：\u002Fcamera\u002Frgb\u002Fimage_raw）\n- **image_reliability**: 图像话题的可靠性：0=系统默认，1=可靠，2=尽力而为（默认：1）\n- **input_depth_topic**: 清晰度图像的相机话题（默认：\u002Fcamera\u002Fdepth\u002Fimage_raw）\n- **depth_image_reliability**: 深度图像话题的可靠性：0=系统默认，1=可靠，2=尽力而为（默认：1）\n- **input_depth_info_topic**: 深度信息的相机话题（默认：\u002Fcamera\u002Fdepth\u002Fcamera_info）\n- **depth_info_reliability**: 深度信息话题的可靠性：0=系统默认，1=可靠，2=尽力而为（默认：1）\n- **target_frame**: 用于变换 3D 框架的坐标系（默认：base_link）\n- **depth_image_units_divisor**: 将深度图像转换为米的除数。取决于所使用的相机（默认：1000）\n- **use_tracking**: 检测后是否激活跟踪功能（默认：True）\n- **use_3d**: 是否激活 3D 检测功能（默认：False）\n- **use_debug**: 是否激活调试节点（默认：True）\n\n## 生命周期节点\n\n之前的更新为包中所有节点添加了生命周期节点支持。该实现尝试通过仅在活动状态加载模型并激活订阅者来减少未配置和非活动状态下的工作负载。\n\n以下是在 30fps 视频流上使用默认 yolov8m.pt 模型的一些资源比较。\n\n| 状态    | CPU 使用率 (i7 12th Gen) | VRAM 使用量 | 带宽使用量 |\n| -------- | ----------------------- | ---------- | --------------- |\n| 活动   | 单核占用 40-50%      | 628 MB     | 最高 200 Mbps  |\n| 非活动 | 单核占用 ~5-7%       | 338 MB     | 0-20 Kbps       |\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmgonzs13_yolo_ros_readme_f75e422ced5a.png\" width=\"100%\" \u002F>\n\u003C\u002Fp>\n\n## 演示\n\n## 物体检测\n\n这是 yolo_ros 的标准行为，包含物体跟踪功能。\n\n```shell\nros2 launch yolo_bringup yolo.launch.py\n```\n\n[![](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmgonzs13_yolo_ros_readme_967a3dbe83d2.png)](https:\u002F\u002Fdrive.google.com\u002Ffile\u002Fd\u002F1gTQt6soSIq1g2QmK7locHDiZ-8MqVl2w\u002Fview?usp=sharing)\n\n## 实例分割\n\n实例掩码是检测到的物体的边界，而不是掩码内部的所有像素。\n\n```shell\nros2 launch yolo_bringup yolo.launch.py model:=yolov8m-seg.pt\n```\n\n[![](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmgonzs13_yolo_ros_readme_02f9abf4bf5d.png)](https:\u002F\u002Fdrive.google.com\u002Ffile\u002Fd\u002F1dwArjDLSNkuOGIB0nSzZR6ABIOCJhAFq\u002Fview?usp=sharing)\n\n## 人体姿态\n\n检测到可见的人及其骨骼关键点。\n\n```shell\nros2 launch yolo_bringup yolo.launch.py model:=yolov8m-pose.pt\n```\n\n[![](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmgonzs13_yolo_ros_readme_8a487dd3501e.png)](https:\u002F\u002Fdrive.google.com\u002Ffile\u002Fd\u002F1pRy9lLSXiFEVFpcbesMCzmTMEoUXGWgr\u002Fview?usp=sharing)\n\n## 3D 物体检测\n\n3D 边界框是通过使用 2D 边界框从 RGB-D 相机的深度图像数据中筛选出来的。只有具有 3D 边界框的物体才会在 2D 图像中显示。\n\n```shell\nros2 launch yolo_bringup yolo.launch.py use_3d:=True\n```\n\n[![](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmgonzs13_yolo_ros_readme_4e1d9c06578e.png)](https:\u002F\u002Fdrive.google.com\u002Ffile\u002Fd\u002F1ZcN_u9RB9_JKq37mdtpzXx3b44tlU-pr\u002Fview?usp=sharing)\n\n## 3D 物体检测（使用实例分割掩码）\n\n在此过程中，深度图像数据会根据实例掩码获取的最大和最小值进行筛选。只有具有 3D 边界框的物体才会在 2D 图像中显示。\n\n```shell\nros2 launch yolo_bringup yolo.launch.py model:=yolov8m-seg.pt use_3d:=True\n```\n\n[![](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmgonzs13_yolo_ros_readme_e1ec8b052fc6.png)](https:\u002F\u002Fdrive.google.com\u002Ffile\u002Fd\u002F1wVZgi5GLkAYxv3GmTxX5z-vB8RQdwqLP\u002Fview?usp=sharing)\n\n## 3D 人体姿态\n\n每个关键点都会投影到深度图像中，并用紫色球体表示。只有具有 3D 边界框的物体才会在 2D 图像中显示。\n\n```shell\nros2 launch yolo_bringup yolo.launch.py model:=yolov8m-pose.pt use_3d:=True\n```\n\n[![](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmgonzs13_yolo_ros_readme_61fea9f07e43.png)](https:\u002F\u002Fdrive.google.com\u002Ffile\u002Fd\u002F1j4VjCAsOCx_mtM2KFPOLkpJogM0t227r\u002Fview?usp=sharing)","# yolo_ros 快速上手指南\n\n`yolo_ros` 是一个基于 ROS 2 的封装工具，集成了 Ultralytics 的 YOLO 系列模型。它支持目标检测、跟踪、实例分割、人体姿态估计以及带方向的目标框（OBB），并提供了基于深度图像的 3D 检测功能。\n\n## 1. 环境准备\n\n在开始之前，请确保您的开发环境满足以下要求：\n\n*   **操作系统**: Ubuntu 20.04\u002F22.04\u002F24.04 (取决于 ROS 2 版本)\n*   **ROS 2 版本**: 支持 Humble, Iron, Jazzy, Kilted, Rolling 等主流发行版。\n*   **Python**: Python 3.x (通常随 ROS 2 安装)\n*   **GPU 加速 (可选)**: 如需使用 CUDA 加速，请确保已安装 NVIDIA 驱动及对应的 CUDA Toolkit。\n*   **工作空间**: 拥有一个标准的 ROS 2 工作空间 (例如 `~\u002Fros2_ws`)。\n\n> **提示**: 如果您在中国大陆地区，建议在安装依赖时配置国内镜像源（如清华源或阿里源）以加速下载。\n\n## 2. 安装步骤\n\n请在您的 ROS 2 工作空间源码目录下执行以下命令：\n\n```shell\n# 进入工作空间源码目录\ncd ~\u002Fros2_ws\u002Fsrc\n\n# 克隆仓库\ngit clone https:\u002F\u002Fgithub.com\u002Fmgonzs13\u002Fyolo_ros.git\n\n# 安装 Python 依赖\n# 建议先配置 pip 国内镜像加速，例如：\n# pip3 install -r yolo_ros\u002Frequirements.txt -i https:\u002F\u002Fpypi.tuna.tsinghua.edu.cn\u002Fsimple\npip3 install -r yolo_ros\u002Frequirements.txt\n\n# 返回工作空间根目录\ncd ~\u002Fros2_ws\n\n# 安装系统依赖\nrosdep install --from-paths src --ignore-src -r -y\n\n# 编译项目\ncolcon build\n\n# 加载环境变量\nsource ~\u002Fros2_ws\u002Finstall\u002Fsetup.bash\n```\n\n### Docker 方式 (可选)\n\n如果您希望使用容器化部署，特别是为了隔离环境或使用 GPU：\n\n```shell\n# 构建镜像\ndocker build -t yolo_ros .\n\n# 运行容器 (如需启用 GPU，请确保已安装 NVIDIA Container Toolkit 并添加 --gpus all)\ndocker run -it --rm --gpus all yolo_ros\n```\n\n## 3. 基本使用\n\n`yolo_ros` 提供了针对不同 YOLO 版本的启动文件。以下是最简单的启动示例。\n\n### 启动默认模型 (YOLOv8)\n\n默认情况下，工具会加载 `yolov8m.pt` 模型并进行目标检测与跟踪。请确保您有一个摄像头话题发布 RGB 图像（默认为 `\u002Fcamera\u002Frgb\u002Fimage_raw`）。\n\n```shell\nros2 launch yolo_bringup yolov8.launch.py\n```\n\n### 切换其他模型\n\n您可以通过修改启动参数轻松切换不同的 YOLO 版本或任务类型：\n\n*   **YOLOv5**:\n    ```shell\n    ros2 launch yolo_bringup yolov5.launch.py\n    ```\n*   **实例分割 (Segmentation)**:\n    ```shell\n    ros2 launch yolo_bringup yolo.launch.py model:=yolov8m-seg.pt\n    ```\n*   **人体姿态估计 (Pose)**:\n    ```shell\n    ros2 launch yolo_bringup yolo.launch.py model:=yolov8m-pose.pt\n    ```\n*   **YOLO-World (开放词汇检测)**:\n    ```shell\n    ros2 launch yolo_bringup yolo-world.launch.py\n    ```\n\n### 核心话题说明\n\n启动后，您可以关注以下主要话题：\n\n*   `\u002Fyolo\u002Fdetections`: 检测结果（包含边界框、类别、置信度）。\n*   `\u002Fyolo\u002Ftracking`: 带有追踪 ID 的目标结果。\n*   `\u002Fyolo\u002Fdebug_image`: 可视化调试图像，可直接在 `rviz2` 中查看。\n*   `\u002Fyolo\u002Fdetections_3d`: 如果启用了深度相机并设置 `use_3d:=true`，将输出 3D 检测结果。\n\n### 常用参数调整\n\n您可以在启动命令中通过 `:=` 覆盖默认参数，例如指定设备或调整阈值：\n\n```shell\n# 指定使用 CPU 运行，并将检测阈值设为 0.6\nros2 launch yolo_bringup yolov8.launch.py device:=cpu threshold:=0.6\n```\n\n主要可用参数包括：\n*   `model`: 模型权重文件路径或名称 (默认: `yolov8m.pt`)\n*   `device`: 推理设备 (默认: `cuda:0`, 可选 `cpu`)\n*   `threshold`: 检测置信度阈值 (默认: `0.5`)\n*   `input_image_topic`: 输入图像话题 (默认: `\u002Fcamera\u002Frgb\u002Fimage_raw`)\n*   `use_3d`: 是否启用 3D 检测 (默认: `False`)","某仓储物流团队正在开发一款基于 ROS 2 的自主移动机器人，需要在动态环境中实时识别货架、托盘及工作人员以规划安全路径。\n\n### 没有 yolo_ros 时\n- **集成繁琐**：开发者需手动编写复杂的桥接代码，将摄像头图像从 ROS 2 话题转换给独立的 YOLO 推理脚本，再回传结果，通信延迟高且易出错。\n- **功能单一**：仅能实现基础的 2D 框检测，若要获取障碍物精确距离或进行人体姿态分析以防碰撞，需额外融合深度相机数据并自行开发 3D 算法。\n- **维护困难**：YOLO 模型升级（如从 v8 迁至 v11）需重构大量接口代码，且难以直接复用 Ultralytics 官方最新的跟踪与分割特性。\n- **调试低效**：缺乏标准化的 ROS 2 节点封装，无法直接使用 Rviz2 直观可视化检测结果，排查感知故障耗时费力。\n\n### 使用 yolo_ros 后\n- **即插即用**：直接加载预配置的 ROS 2 节点，订阅图像话题即可输出标准的检测消息，无缝支持 Humble 至 Jazzy 多个发行版，开发周期缩短数天。\n- **多维感知**：一键开启 3D 目标检测与人姿估计模式，利用深度图自动解算障碍物空间坐标，显著提升机器人在人流密集区的避障安全性。\n- **持续演进**：平滑支持 YOLOv8 到 v12 全系列模型，无需修改业务逻辑即可享受官方最新的实例分割与旋转框（OBB）能力，适应不同货物形态。\n- **可视可控**：原生兼容 Rviz2 显示检测框与骨架点，结合 ROS 2 参数动态调整置信度阈值，现场调试效率大幅提升。\n\nyolo_ros 通过标准化封装消除了算法与机器人系统间的隔阂，让开发者能专注于上层应用逻辑而非底层集成琐事。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmgonzs13_yolo_ros_967a3dbe.jpg","mgonzs13","Miguel Ángel González Santamarta","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002Fmgonzs13_74110317.png","Ph.D. at Universidad de León. Research in Robotics and Artificial Intelligence.","University of León","León, Spain","mgons@unileon.es","miggsant",null,"https:\u002F\u002Fgithub.com\u002Fmgonzs13",[87,91,95],{"name":88,"color":89,"percentage":90},"Python","#3572A5",98.5,{"name":92,"color":93,"percentage":94},"Dockerfile","#384d54",0.9,{"name":96,"color":97,"percentage":98},"CMake","#DA3434",0.6,1000,238,"2026-04-03T07:23:20","GPL-3.0","Linux","可选但推荐（用于加速）。若使用 CUDA，需安装 NVIDIA Container Toolkit 并指定 --gpus all。默认设备参数为 'cuda:0'，支持半精度推理 (FP16)。未明确具体显存大小要求，但生命周期节点测试显示激活状态下显存占用约 628MB (基于 yolov8m 模型)。","未说明",{"notes":107,"python":108,"dependencies":109},"该工具是 ROS 2 包，主要支持 Linux 环境（提供 Humble 到 Rolling 多个发行版的 Docker 镜像）。支持从 YOLOv3 到 YOLOv12 以及 YOLO-World\u002FYOLOE 等多种模型。具备生命周期节点 (Lifecycle Nodes) 功能，可在非激活状态下显著降低 CPU 和显存占用。若需运行 3D 检测功能，需提供深度图像话题 (depth images)。默认输入图像编码为 bgr8。","未说明 (通过 requirements.txt 安装依赖)",[110,111,112,113,114],"ROS 2 (Humble, Iron, Jazzy, Kilted, Rolling)","Ultralytics (YOLO 模型库)","colcon","rosdep","NVIDIA Container Toolkit (如需 GPU 加速)",[14,37],[117,118,119,120,121,122,123,124,125,126,127,128,129,130,131,132,133],"object-detection","ros2","ultralytics","yolov8","object-tracking","3d-human-pose-estimation","3d-object-detection","human-pose-estimation","instance-segmentation","yolov9","yolov10","obb","oriented-bounding-box","yolov11","yolov12","yoloe","yolo","2026-03-27T02:49:30.150509","2026-04-06T07:15:08.115177",[],[138,143,148,153,158,163,168,173,178,183,188,193,198,203,208,213,218,223,227,232],{"id":139,"version":140,"summary_zh":141,"released_at":142},107121,"4.5.1","### Changelog from version 4.5.0 to 4.5.1:\n4b2796c new version 4.5.1\nd003178 Upgrading ultralytics + adding YOLOv26 launch","2026-01-20T08:30:08",{"id":144,"version":145,"summary_zh":146,"released_at":147},107122,"4.5.0","### Changelog from version 4.4.1 to 4.5.0:\n7ef13c9 new version 4.5.0\nb918bef Fixing Python format\n8d20df2 Add robust float validation to prevent crashes in depth processing (#106)\n7df174d Fix lifecycle teardown in tracking node (#107)\n193c545 Moving yolo model to device on activate\nd55009f Updating doxygen deployment\n3076c1e New compute functions for 3D detection\n2bec8d8 Avoiding isfinite in 3D detection\n457fe31 Adding docstring to class methods\n5fd5330 Removing maximum_detection_threshold","2026-01-19T21:11:42",{"id":149,"version":150,"summary_zh":151,"released_at":152},107123,"4.4.1","### Changelog from version 4.4.0 to 4.4.1:\ndf71824 new version 4.4.1\n13f496f Add missing dependencies to package.xml","2026-01-04T14:23:57",{"id":154,"version":155,"summary_zh":156,"released_at":157},107124,"4.4.0","### Changelog from version 4.3.1 to 4.4.0:\naaf1b54 new version 4.4.0\nb69ed0d Adding fuse_model doc in README\n171a467 Fixing on_shutdown lifecycle methods\nff85f6b Adding fuse_model param to control if fusing the yolo model (default is False)\n60e7c5e adding the redirect index to the root dir of gh-pages\n22f50dd removing Doxygen from distro table\n6d861eb Optimizing Dockerfile\nb03f3f8 fixing license name","2025-12-22T10:53:39",{"id":159,"version":160,"summary_zh":161,"released_at":162},107125,"4.3.1","### Changelog from version 4.3.0 to 4.3.1:\necdc718 new version 4.3.1\n8dc6320 removing foxy and galactic\n0ec5e18 fixing dependencies","2025-08-25T09:05:50",{"id":164,"version":165,"summary_zh":166,"released_at":167},107126,"4.3.0","### Changelog from version 4.2.0 to 4.3.0:\n0ed149e new version 4.3.0\n59f8980 Add YOLOE Support (#98)\nc811121 Ros2 fixes (#95)\n5936e97 adding foxy and galactic workflows\n6c9f76d adding workflows for iron, kilted and rolling","2025-08-25T08:51:29",{"id":169,"version":170,"summary_zh":171,"released_at":172},107127,"4.2.0","### Changelog from version 4.1.1 to 4.2.0:\nacc12c7 new version 4.2.0\nc1712e6 adding workflows for releases\n8a2d743 yolo_encoding param added\na52adf7 upgrading lap dependency\naa7eb07 cleaning bringup package\na99a101 debug_node: Use the input message header as header of the debug message (#84)","2025-04-10T20:19:09",{"id":174,"version":175,"summary_zh":176,"released_at":177},107128,"4.1.1","### Changelog from version 4.1.0 to 4.1.1:\n36bbce0 new version 4.1.1\n245b19e adding badges to README\n9296b6e fixing default params values\nbcba7f8 ultralytics updated\nbf5d5b6 default image_reliability value set to 1\nb181d7a adding create release workflow","2025-03-17T21:14:25",{"id":179,"version":180,"summary_zh":181,"released_at":182},107129,"4.1.0","  - Fixing workflows names\r\n  - Fixing image encoding\r\n  - YOLO-NAS removed\r\n  - YOLOv12 added\r\n  - minor fixes to format","2025-02-21T15:54:03",{"id":184,"version":185,"summary_zh":186,"released_at":187},107130,"4.0.1","  - minor fixes in README\r\n  - try\u002Fexcept for model file not existing and model fuse call","2024-11-15T21:04:42",{"id":189,"version":190,"summary_zh":191,"released_at":192},107131,"4.0.0","1. **Renaming**: Changed `yolov8_ros` to `yolo_ros` for a more consistent naming convention.\r\n2. **New Features**:\r\n   - Added `yolo-world` node for demonstration purposes.\r\n   - Created a Dockerfile for containerized deployments.\r\n3. **Inference Parameters**: Added new parameters to the inference functions to allow for customized usage.\r\n4. **Launch File Update**: Renamed `yolo-base.launch.py` to `yolo.launch.py`.\r\n5. **CI Enhancements**:\r\n   - Implemented GitHub Actions workflows for continuous integration and testing.\r\n   - CI includes formatting checks and docker build.\r\n6. **Formatting and Code Quality**:\r\n   - Code has been formatted using `black` for consistency.\r\n   - Additional minor fixes to ensure a clean codebase.\r\n7. **Documentation Updated**:\r\n   - New inference parameters added to README\r\n   - New launch files added to README","2024-10-31T22:18:58",{"id":194,"version":195,"summary_zh":196,"released_at":197},107132,"3.4.0","  - yolov11 added\r\n  - yolov5 launch example added\r\n  - docs for yolo models included in README\r\n  - ultralytics 8.3.3","2024-10-02T18:53:26",{"id":199,"version":200,"summary_zh":201,"released_at":202},107133,"3.3.3","  - check hypothesis, boxes, masks and keypoints before creating the msgs","2024-09-03T11:11:12",{"id":204,"version":205,"summary_zh":206,"released_at":207},107134,"3.3.2","  - fixing logs for lifecycle nodes","2024-08-29T08:44:07",{"id":209,"version":210,"summary_zh":211,"released_at":212},107135,"3.3.1","   - fuse for YOLOv10\r\n   - typing for enable_cb in yolov8_node\r\n   - ultralytics updated","2024-08-21T08:47:37",{"id":214,"version":215,"summary_zh":216,"released_at":217},107136,"3.3.0","  - YOLO-NAS models added","2024-08-02T08:39:23",{"id":219,"version":220,"summary_zh":221,"released_at":222},107137,"3.2.0","OBB supported","2024-07-22T14:45:09",{"id":224,"version":225,"summary_zh":127,"released_at":226},107138,"3.1.1","2024-06-20T20:56:11",{"id":228,"version":229,"summary_zh":230,"released_at":231},107139,"3.1.0","Using 2D mask to create 3D bounding boxes","2024-06-05T13:28:16",{"id":233,"version":234,"summary_zh":235,"released_at":236},107140,"3.0.2","Color 3D markers of 3D bounding boxes","2024-04-30T10:01:25"]