[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"similar-open-mmlab--mmdetection3d":3,"tool-open-mmlab--mmdetection3d":61},[4,18,26,36,44,53],{"id":5,"name":6,"github_repo":7,"description_zh":8,"stars":9,"difficulty_score":10,"last_commit_at":11,"category_tags":12,"status":17},4358,"openclaw","openclaw\u002Fopenclaw","OpenClaw 是一款专为个人打造的本地化 AI 助手，旨在让你在自己的设备上拥有完全可控的智能伙伴。它打破了传统 AI 助手局限于特定网页或应用的束缚，能够直接接入你日常使用的各类通讯渠道，包括微信、WhatsApp、Telegram、Discord、iMessage 等数十种平台。无论你在哪个聊天软件中发送消息，OpenClaw 都能即时响应，甚至支持在 macOS、iOS 和 Android 设备上进行语音交互，并提供实时的画布渲染功能供你操控。\n\n这款工具主要解决了用户对数据隐私、响应速度以及“始终在线”体验的需求。通过将 AI 部署在本地，用户无需依赖云端服务即可享受快速、私密的智能辅助，真正实现了“你的数据，你做主”。其独特的技术亮点在于强大的网关架构，将控制平面与核心助手分离，确保跨平台通信的流畅性与扩展性。\n\nOpenClaw 非常适合希望构建个性化工作流的技术爱好者、开发者，以及注重隐私保护且不愿被单一生态绑定的普通用户。只要具备基础的终端操作能力（支持 macOS、Linux 及 Windows WSL2），即可通过简单的命令行引导完成部署。如果你渴望拥有一个懂你",349277,3,"2026-04-06T06:32:30",[13,14,15,16],"Agent","开发框架","图像","数据工具","ready",{"id":19,"name":20,"github_repo":21,"description_zh":22,"stars":23,"difficulty_score":10,"last_commit_at":24,"category_tags":25,"status":17},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,"2026-04-05T11:01:52",[14,15,13],{"id":27,"name":28,"github_repo":29,"description_zh":30,"stars":31,"difficulty_score":32,"last_commit_at":33,"category_tags":34,"status":17},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",142651,2,"2026-04-06T23:34:12",[14,13,35],"语言模型",{"id":37,"name":38,"github_repo":39,"description_zh":40,"stars":41,"difficulty_score":32,"last_commit_at":42,"category_tags":43,"status":17},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。",107888,"2026-04-06T11:32:50",[14,15,13],{"id":45,"name":46,"github_repo":47,"description_zh":48,"stars":49,"difficulty_score":32,"last_commit_at":50,"category_tags":51,"status":17},4721,"markitdown","microsoft\u002Fmarkitdown","MarkItDown 是一款由微软 AutoGen 团队打造的轻量级 Python 工具，专为将各类文件高效转换为 Markdown 格式而设计。它支持 PDF、Word、Excel、PPT、图片（含 OCR）、音频（含语音转录）、HTML 乃至 YouTube 链接等多种格式的解析，能够精准提取文档中的标题、列表、表格和链接等关键结构信息。\n\n在人工智能应用日益普及的今天，大语言模型（LLM）虽擅长处理文本，却难以直接读取复杂的二进制办公文档。MarkItDown 恰好解决了这一痛点，它将非结构化或半结构化的文件转化为模型“原生理解”且 Token 效率极高的 Markdown 格式，成为连接本地文件与 AI 分析 pipeline 的理想桥梁。此外，它还提供了 MCP（模型上下文协议）服务器，可无缝集成到 Claude Desktop 等 LLM 应用中。\n\n这款工具特别适合开发者、数据科学家及 AI 研究人员使用，尤其是那些需要构建文档检索增强生成（RAG）系统、进行批量文本分析或希望让 AI 助手直接“阅读”本地文件的用户。虽然生成的内容也具备一定可读性，但其核心优势在于为机器",93400,"2026-04-06T19:52:38",[52,14],"插件",{"id":54,"name":55,"github_repo":56,"description_zh":57,"stars":58,"difficulty_score":10,"last_commit_at":59,"category_tags":60,"status":17},4487,"LLMs-from-scratch","rasbt\u002FLLMs-from-scratch","LLMs-from-scratch 是一个基于 PyTorch 的开源教育项目，旨在引导用户从零开始一步步构建一个类似 ChatGPT 的大型语言模型（LLM）。它不仅是同名技术著作的官方代码库，更提供了一套完整的实践方案，涵盖模型开发、预训练及微调的全过程。\n\n该项目主要解决了大模型领域“黑盒化”的学习痛点。许多开发者虽能调用现成模型，却难以深入理解其内部架构与训练机制。通过亲手编写每一行核心代码，用户能够透彻掌握 Transformer 架构、注意力机制等关键原理，从而真正理解大模型是如何“思考”的。此外，项目还包含了加载大型预训练权重进行微调的代码，帮助用户将理论知识延伸至实际应用。\n\nLLMs-from-scratch 特别适合希望深入底层原理的 AI 开发者、研究人员以及计算机专业的学生。对于不满足于仅使用 API，而是渴望探究模型构建细节的技术人员而言，这是极佳的学习资源。其独特的技术亮点在于“循序渐进”的教学设计：将复杂的系统工程拆解为清晰的步骤，配合详细的图表与示例，让构建一个虽小但功能完备的大模型变得触手可及。无论你是想夯实理论基础，还是为未来研发更大规模的模型做准备",90106,"2026-04-06T11:19:32",[35,15,13,14],{"id":62,"github_repo":63,"name":64,"description_en":65,"description_zh":66,"ai_summary_zh":66,"readme_en":67,"readme_zh":68,"quickstart_zh":69,"use_case_zh":70,"hero_image_url":71,"owner_login":72,"owner_name":73,"owner_avatar_url":74,"owner_bio":75,"owner_company":76,"owner_location":76,"owner_email":76,"owner_twitter":73,"owner_website":77,"owner_url":78,"languages":79,"stars":103,"forks":104,"last_commit_at":105,"license":106,"difficulty_score":107,"env_os":108,"env_gpu":109,"env_ram":110,"env_deps":111,"category_tags":118,"github_topics":120,"view_count":32,"oss_zip_url":76,"oss_zip_packed_at":76,"status":17,"created_at":125,"updated_at":126,"faqs":127,"releases":156},4985,"open-mmlab\u002Fmmdetection3d","mmdetection3d","OpenMMLab's next-generation platform for general 3D object detection.","MMDetection3D 是 OpenMMLab 推出的新一代通用 3D 物体检测开源平台，旨在为自动驾驶、机器人感知等场景提供强大的三维视觉分析能力。它核心解决了从激光雷达点云、多视角图像或融合数据中，精准识别并定位三维空间内物体（如车辆、行人）的技术难题，让机器能够像人类一样理解立体环境。\n\n这款工具特别适合计算机视觉领域的研究人员、算法工程师以及高校开发者使用。无论是希望复现前沿论文模型，还是需要在实际项目中快速部署 3D 检测功能，MMDetection3D 都能提供坚实支持。其独特亮点在于高度模块化的架构设计，内置了丰富的主流算法库（如 PointPillars、MVX-Net 等）和标准数据集接口。用户无需从零编写底层代码，即可通过简洁的配置灵活组合网络组件，轻松实现新方法的验证与对比。凭借完善的文档和社区支持，MMDetection3D 显著降低了 3D 视觉技术的入门门槛，是推动三维感知技术从理论研究走向产业落地的得力助手。","\u003Cdiv align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fopen-mmlab_mmdetection3d_readme_bc52b0de66dc.png\" width=\"600\"\u002F>\n  \u003Cdiv> \u003C\u002Fdiv>\n  \u003Cdiv align=\"center\">\n    \u003Cb>\u003Cfont size=\"5\">OpenMMLab website\u003C\u002Ffont>\u003C\u002Fb>\n    \u003Csup>\n      \u003Ca href=\"https:\u002F\u002Fopenmmlab.com\">\n        \u003Ci>\u003Cfont size=\"4\">HOT\u003C\u002Ffont>\u003C\u002Fi>\n      \u003C\u002Fa>\n    \u003C\u002Fsup>\n        \n    \u003Cb>\u003Cfont size=\"5\">OpenMMLab platform\u003C\u002Ffont>\u003C\u002Fb>\n    \u003Csup>\n      \u003Ca href=\"https:\u002F\u002Fplatform.openmmlab.com\">\n        \u003Ci>\u003Cfont size=\"4\">TRY IT OUT\u003C\u002Ffont>\u003C\u002Fi>\n      \u003C\u002Fa>\n    \u003C\u002Fsup>\n  \u003C\u002Fdiv>\n  \u003Cdiv> \u003C\u002Fdiv>\n\n[![PyPI](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fv\u002Fmmdet3d)](https:\u002F\u002Fpypi.org\u002Fproject\u002Fmmdet3d)\n[![docs](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdocs-latest-blue)](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002F)\n[![badge](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fworkflows\u002Fbuild\u002Fbadge.svg)](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Factions)\n[![codecov](https:\u002F\u002Fcodecov.io\u002Fgh\u002Fopen-mmlab\u002Fmmdetection3d\u002Fbranch\u002Fmain\u002Fgraph\u002Fbadge.svg)](https:\u002F\u002Fcodecov.io\u002Fgh\u002Fopen-mmlab\u002Fmmdetection3d)\n[![license](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Flicense\u002Fopen-mmlab\u002Fmmdetection3d.svg)](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fblob\u002Fmain\u002FLICENSE)\n[![open issues](https:\u002F\u002Fisitmaintained.com\u002Fbadge\u002Fopen\u002Fopen-mmlab\u002Fmmdetection3d.svg)](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fissues)\n[![issue resolution](https:\u002F\u002Fisitmaintained.com\u002Fbadge\u002Fresolution\u002Fopen-mmlab\u002Fmmdetection3d.svg)](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fissues)\n\n[📘Documentation](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002F) |\n[🛠️Installation](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fget_started.html) |\n[👀Model Zoo](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fmodel_zoo.html) |\n[🆕Update News](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fnotes\u002Fchangelog.html) |\n[🚀Ongoing Projects](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fprojects) |\n[🤔Reporting Issues](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fissues\u002Fnew\u002Fchoose)\n\n\u003C\u002Fdiv>\n\n\u003Cdiv align=\"center\">\n\nEnglish | [简体中文](README_zh-CN.md)\n\n\u003C\u002Fdiv>\n\n\u003Cdiv align=\"center\">\n  \u003Ca href=\"https:\u002F\u002Fopenmmlab.medium.com\u002F\" style=\"text-decoration:none;\">\n    \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fopen-mmlab_mmdetection3d_readme_062337b0e5ec.png\" width=\"3%\" alt=\"\" \u002F>\u003C\u002Fa>\n  \u003Cimg src=\"https:\u002F\u002Fuser-images.githubusercontent.com\u002F25839884\u002F218346358-56cc8e2f-a2b8-487f-9088-32480cceabcf.png\" width=\"3%\" alt=\"\" \u002F>\n  \u003Ca href=\"https:\u002F\u002Fdiscord.com\u002Fchannels\u002F1037617289144569886\u002F1046608014234370059\" style=\"text-decoration:none;\">\n    \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fopen-mmlab_mmdetection3d_readme_6342e5371027.png\" width=\"3%\" alt=\"\" \u002F>\u003C\u002Fa>\n  \u003Cimg src=\"https:\u002F\u002Fuser-images.githubusercontent.com\u002F25839884\u002F218346358-56cc8e2f-a2b8-487f-9088-32480cceabcf.png\" width=\"3%\" alt=\"\" \u002F>\n  \u003Ca href=\"https:\u002F\u002Ftwitter.com\u002FOpenMMLab\" style=\"text-decoration:none;\">\n    \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fopen-mmlab_mmdetection3d_readme_04c3beda0b07.png\" width=\"3%\" alt=\"\" \u002F>\u003C\u002Fa>\n  \u003Cimg src=\"https:\u002F\u002Fuser-images.githubusercontent.com\u002F25839884\u002F218346358-56cc8e2f-a2b8-487f-9088-32480cceabcf.png\" width=\"3%\" alt=\"\" \u002F>\n  \u003Ca href=\"https:\u002F\u002Fwww.youtube.com\u002Fopenmmlab\" style=\"text-decoration:none;\">\n    \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fopen-mmlab_mmdetection3d_readme_204fe79b5a90.png\" width=\"3%\" alt=\"\" \u002F>\u003C\u002Fa>\n  \u003Cimg src=\"https:\u002F\u002Fuser-images.githubusercontent.com\u002F25839884\u002F218346358-56cc8e2f-a2b8-487f-9088-32480cceabcf.png\" width=\"3%\" alt=\"\" \u002F>\n  \u003Ca href=\"https:\u002F\u002Fspace.bilibili.com\u002F1293512903\" style=\"text-decoration:none;\">\n    \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fopen-mmlab_mmdetection3d_readme_8655b6233577.png\" width=\"3%\" alt=\"\" \u002F>\u003C\u002Fa>\n  \u003Cimg src=\"https:\u002F\u002Fuser-images.githubusercontent.com\u002F25839884\u002F218346358-56cc8e2f-a2b8-487f-9088-32480cceabcf.png\" width=\"3%\" alt=\"\" \u002F>\n  \u003Ca href=\"https:\u002F\u002Fwww.zhihu.com\u002Fpeople\u002Fopenmmlab\" style=\"text-decoration:none;\">\n    \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fopen-mmlab_mmdetection3d_readme_447c4737c11f.png\" width=\"3%\" alt=\"\" \u002F>\u003C\u002Fa>\n\u003C\u002Fdiv>\n\n## Introduction\n\nMMDetection3D is an open source object detection toolbox based on PyTorch, towards the next-generation platform for general 3D detection. It is a part of the [OpenMMLab](https:\u002F\u002Fopenmmlab.com\u002F) project.\n\nThe main branch works with **PyTorch 1.8+**.\n\n![demo image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fopen-mmlab_mmdetection3d_readme_5e593c5a783a.gif)\n\n\u003Cdetails open>\n\u003Csummary>Major features\u003C\u002Fsummary>\n\n- **Support multi-modality\u002Fsingle-modality detectors out of box**\n\n  It directly supports multi-modality\u002Fsingle-modality detectors including MVXNet, VoteNet, PointPillars, etc.\n\n- **Support indoor\u002Foutdoor 3D detection out of box**\n\n  It directly supports popular indoor and outdoor 3D detection datasets, including ScanNet, SUNRGB-D, Waymo, nuScenes, Lyft, and KITTI. For nuScenes dataset, we also support [nuImages dataset](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Ftree\u002Fmain\u002Fconfigs\u002Fnuimages).\n\n- **Natural integration with 2D detection**\n\n  All the about **300+ models, methods of 40+ papers**, and modules supported in [MMDetection](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection\u002Fblob\u002F3.x\u002Fdocs\u002Fen\u002Fmodel_zoo.md) can be trained or used in this codebase.\n\n- **High efficiency**\n\n  It trains faster than other codebases. The main results are as below. Details can be found in [benchmark.md](.\u002Fdocs\u002Fen\u002Fnotes\u002Fbenchmarks.md). We compare the number of samples trained per second (the higher, the better). The models that are not supported by other codebases are marked by `✗`.\n\n  |       Methods       | MMDetection3D | [OpenPCDet](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002FOpenPCDet) | [votenet](https:\u002F\u002Fgithub.com\u002Ffacebookresearch\u002Fvotenet) | [Det3D](https:\u002F\u002Fgithub.com\u002Fpoodarchu\u002FDet3D) |\n  | :-----------------: | :-----------: | :--------------------------------------------------: | :----------------------------------------------------: | :-----------------------------------------: |\n  |       VoteNet       |      358      |                          ✗                           |                           77                           |                      ✗                      |\n  |  PointPillars-car   |      141      |                          ✗                           |                           ✗                            |                     140                     |\n  | PointPillars-3class |      107      |                          44                          |                           ✗                            |                      ✗                      |\n  |       SECOND        |      40       |                          30                          |                           ✗                            |                      ✗                      |\n  |       Part-A2       |      17       |                          14                          |                           ✗                            |                      ✗                      |\n\n\u003C\u002Fdetails>\n\nLike [MMDetection](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection) and [MMCV](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmcv), MMDetection3D can also be used as a library to support different projects on top of it.\n\n## What's New\n\n### Highlight\n\nIn version 1.4, MMDetecion3D refactors the Waymo dataset and accelerates the preprocessing, training\u002Ftesting setup, and evaluation of Waymo dataset. We also extends the support for camera-based, such as Monocular and BEV, 3D object detection models on Waymo. A detailed description of the Waymo data information is provided [here](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fadvanced_guides\u002Fdatasets\u002Fwaymo.html).\n\nBesides, in version 1.4, MMDetection3D provides [Waymo-mini](https:\u002F\u002Fdownload.openmmlab.com\u002Fmmdetection3d\u002Fdata\u002Fwaymo_mmdet3d_after_1x4\u002Fwaymo_mini.tar.gz) to help community users get started with Waymo and use it for quick iterative development.\n\n**v1.4.0** was released in 8\u002F1\u002F2024：\n\n- Support the training of [DSVT](\u003C(https:\u002F\u002Farxiv.org\u002Fabs\u002F2301.06051)>) in `projects`\n- Support [Nerf-Det](https:\u002F\u002Farxiv.org\u002Fabs\u002F2307.14620) in `projects`\n- Refactor Waymo dataset\n\n**v1.3.0** was released in 18\u002F10\u002F2023:\n\n- Support [CENet](https:\u002F\u002Farxiv.org\u002Fabs\u002F2207.12691) in `projects`\n- Enhance demos with new 3D inferencers\n\n**v1.2.0** was released in 4\u002F7\u002F2023\n\n- Support [New Config Type](https:\u002F\u002Fmmengine.readthedocs.io\u002Fen\u002Flatest\u002Fadvanced_tutorials\u002Fconfig.html#a-pure-python-style-configuration-file-beta) in `mmdet3d\u002Fconfigs`\n- Support the inference of [DSVT](\u003C(https:\u002F\u002Farxiv.org\u002Fabs\u002F2301.06051)>) in `projects`\n- Support downloading datasets from [OpenDataLab](https:\u002F\u002Fopendatalab.com\u002F) using `mim`\n\n**v1.1.1** was released in 30\u002F5\u002F2023:\n\n- Support [TPVFormer](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2302.07817.pdf) in `projects`\n- Support the training of BEVFusion in `projects`\n- Support lidar-based 3D semantic segmentation benchmark\n\n## Installation\n\nPlease refer to [Installation](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fget_started.html) for installation instructions.\n\n## Getting Started\n\nFor detailed user guides and advanced guides, please refer to our [documentation](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002F):\n\n\u003Cdetails>\n\u003Csummary>User Guides\u003C\u002Fsummary>\n\n- [Train & Test](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fuser_guides\u002Findex.html#train-test)\n  - [Learn about Configs](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fuser_guides\u002Fconfig.html)\n  - [Coordinate System](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fuser_guides\u002Fcoord_sys_tutorial.html)\n  - [Dataset Preparation](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fuser_guides\u002Fdataset_prepare.html)\n  - [Customize Data Pipelines](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fuser_guides\u002Fdata_pipeline.html)\n  - [Test and Train on Standard Datasets](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fuser_guides\u002Ftrain_test.html)\n  - [Inference](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fuser_guides\u002Finference.html)\n  - [Train with Customized Datasets](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fuser_guides\u002Fnew_data_model.html)\n- [Useful Tools](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fuser_guides\u002Findex.html#useful-tools)\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>Advanced Guides\u003C\u002Fsummary>\n\n- [Datasets](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fadvanced_guides\u002Findex.html#datasets)\n  - [KITTI Dataset](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fadvanced_guides\u002Fdatasets\u002Fkitti.html)\n  - [NuScenes Dataset](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fadvanced_guides\u002Fdatasets\u002Fnuscenes.html)\n  - [Lyft Dataset](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fadvanced_guides\u002Fdatasets\u002Flyft.html)\n  - [Waymo Dataset](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fadvanced_guides\u002Fdatasets\u002Fwaymo.html)\n  - [SUN RGB-D Dataset](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fadvanced_guides\u002Fdatasets\u002Fsunrgbd.html)\n  - [ScanNet Dataset](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fadvanced_guides\u002Fdatasets\u002Fscannet.html)\n  - [S3DIS Dataset](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fadvanced_guides\u002Fdatasets\u002Fs3dis.html)\n  - [SemanticKITTI Dataset](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fadvanced_guides\u002Fdatasets\u002Fsemantickitti.html)\n- [Supported Tasks](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fadvanced_guides\u002Findex.html#supported-tasks)\n  - [LiDAR-Based 3D Detection](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fadvanced_guides\u002Fsupported_tasks\u002Flidar_det3d.html)\n  - [Vision-Based 3D Detection](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fadvanced_guides\u002Fsupported_tasks\u002Fvision_det3d.html)\n  - [LiDAR-Based 3D Semantic Segmentation](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fadvanced_guides\u002Fsupported_tasks\u002Flidar_sem_seg3d.html)\n- [Customization](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fadvanced_guides\u002Findex.html#customization)\n  - [Customize Datasets](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fadvanced_guides\u002Fcustomize_dataset.html)\n  - [Customize Models](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fadvanced_guides\u002Fcustomize_models.html)\n  - [Customize Runtime Settings](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fadvanced_guides\u002Fcustomize_runtime.html)\n\n\u003C\u002Fdetails>\n\n## Overview of Benchmark and Model Zoo\n\nResults and models are available in the [model zoo](docs\u002Fen\u002Fmodel_zoo.md).\n\n\u003Cdiv align=\"center\">\n  \u003Cb>Components\u003C\u002Fb>\n\u003C\u002Fdiv>\n\u003Ctable align=\"center\">\n  \u003Ctbody>\n    \u003Ctr align=\"center\" valign=\"bottom\">\n      \u003Ctd>\n        \u003Cb>Backbones\u003C\u002Fb>\n      \u003C\u002Ftd>\n      \u003Ctd>\n        \u003Cb>Heads\u003C\u002Fb>\n      \u003C\u002Ftd>\n      \u003Ctd>\n        \u003Cb>Features\u003C\u002Fb>\n      \u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr valign=\"top\">\n      \u003Ctd>\n      \u003Cul>\n        \u003Cli>\u003Ca href=\"configs\u002Fpointnet2\">PointNet (CVPR'2017)\u003C\u002Fa>\u003C\u002Fli>\n        \u003Cli>\u003Ca href=\"configs\u002Fpointnet2\">PointNet++ (NeurIPS'2017)\u003C\u002Fa>\u003C\u002Fli>\n        \u003Cli>\u003Ca href=\"configs\u002Fregnet\">RegNet (CVPR'2020)\u003C\u002Fa>\u003C\u002Fli>\n        \u003Cli>\u003Ca href=\"configs\u002Fdgcnn\">DGCNN (TOG'2019)\u003C\u002Fa>\u003C\u002Fli>\n        \u003Cli>DLA (CVPR'2018)\u003C\u002Fli>\n        \u003Cli>MinkResNet (CVPR'2019)\u003C\u002Fli>\n        \u003Cli>\u003Ca href=\"configs\u002Fminkunet\">MinkUNet (CVPR'2019)\u003C\u002Fa>\u003C\u002Fli>\n        \u003Cli>\u003Ca href=\"configs\u002Fcylinder3d\">Cylinder3D (CVPR'2021)\u003C\u002Fa>\u003C\u002Fli>\n      \u003C\u002Ful>\n      \u003C\u002Ftd>\n      \u003Ctd>\n      \u003Cul>\n        \u003Cli>\u003Ca href=\"configs\u002Ffree_anchor\">FreeAnchor (NeurIPS'2019)\u003C\u002Fa>\u003C\u002Fli>\n      \u003C\u002Ful>\n      \u003C\u002Ftd>\n      \u003Ctd>\n      \u003Cul>\n        \u003Cli>\u003Ca href=\"configs\u002Fdynamic_voxelization\">Dynamic Voxelization (CoRL'2019)\u003C\u002Fa>\u003C\u002Fli>\n      \u003C\u002Ful>\n      \u003C\u002Ftd>\n    \u003C\u002Ftr>\n\u003C\u002Ftd>\n    \u003C\u002Ftr>\n  \u003C\u002Ftbody>\n\u003C\u002Ftable>\n\n\u003Cdiv align=\"center\">\n  \u003Cb>Architectures\u003C\u002Fb>\n\u003C\u002Fdiv>\n\u003Ctable align=\"center\">\n  \u003Ctbody>\n    \u003Ctr align=\"center\" valign=\"middle\">\n      \u003Ctd>\n        \u003Cb>LiDAR-based 3D Object Detection\u003C\u002Fb>\n      \u003C\u002Ftd>\n      \u003Ctd>\n        \u003Cb>Camera-based 3D Object Detection\u003C\u002Fb>\n      \u003C\u002Ftd>\n      \u003Ctd>\n        \u003Cb>Multi-modal 3D Object Detection\u003C\u002Fb>\n      \u003C\u002Ftd>\n      \u003Ctd>\n        \u003Cb>3D Semantic Segmentation\u003C\u002Fb>\n      \u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr valign=\"top\">\n      \u003Ctd>\n        \u003Cli>\u003Cb>Outdoor\u003C\u002Fb>\u003C\u002Fli>\n        \u003Cul>\n            \u003Cli>\u003Ca href=\"configs\u002Fsecond\">SECOND (Sensor'2018)\u003C\u002Fa>\u003C\u002Fli>\n            \u003Cli>\u003Ca href=\"configs\u002Fpointpillars\">PointPillars (CVPR'2019)\u003C\u002Fa>\u003C\u002Fli>\n            \u003Cli>\u003Ca href=\"configs\u002Fssn\">SSN (ECCV'2020)\u003C\u002Fa>\u003C\u002Fli>\n            \u003Cli>\u003Ca href=\"configs\u002F3dssd\">3DSSD (CVPR'2020)\u003C\u002Fa>\u003C\u002Fli>\n            \u003Cli>\u003Ca href=\"configs\u002Fsassd\">SA-SSD (CVPR'2020)\u003C\u002Fa>\u003C\u002Fli>\n            \u003Cli>\u003Ca href=\"configs\u002Fpoint_rcnn\">PointRCNN (CVPR'2019)\u003C\u002Fa>\u003C\u002Fli>\n            \u003Cli>\u003Ca href=\"configs\u002Fparta2\">Part-A2 (TPAMI'2020)\u003C\u002Fa>\u003C\u002Fli>\n            \u003Cli>\u003Ca href=\"configs\u002Fcenterpoint\">CenterPoint (CVPR'2021)\u003C\u002Fa>\u003C\u002Fli>\n            \u003Cli>\u003Ca href=\"configs\u002Fpv_rcnn\">PV-RCNN (CVPR'2020)\u003C\u002Fa>\u003C\u002Fli>\n            \u003Cli>\u003Ca href=\"projects\u002FCenterFormer\">CenterFormer (ECCV'2022)\u003C\u002Fa>\u003C\u002Fli>\n        \u003C\u002Ful>\n        \u003Cli>\u003Cb>Indoor\u003C\u002Fb>\u003C\u002Fli>\n        \u003Cul>\n            \u003Cli>\u003Ca href=\"configs\u002Fvotenet\">VoteNet (ICCV'2019)\u003C\u002Fa>\u003C\u002Fli>\n            \u003Cli>\u003Ca href=\"configs\u002Fh3dnet\">H3DNet (ECCV'2020)\u003C\u002Fa>\u003C\u002Fli>\n            \u003Cli>\u003Ca href=\"configs\u002Fgroupfree3d\">Group-Free-3D (ICCV'2021)\u003C\u002Fa>\u003C\u002Fli>\n            \u003Cli>\u003Ca href=\"configs\u002Ffcaf3d\">FCAF3D (ECCV'2022)\u003C\u002Fa>\u003C\u002Fli>\n            \u003Cli>\u003Ca href=\"projects\u002FTR3D\">TR3D (ArXiv'2023)\u003C\u002Fa>\u003C\u002Fli>\n      \u003C\u002Ful>\n      \u003C\u002Ftd>\n      \u003Ctd>\n        \u003Cli>\u003Cb>Outdoor\u003C\u002Fb>\u003C\u002Fli>\n        \u003Cul>\n          \u003Cli>\u003Ca href=\"configs\u002Fimvoxelnet\">ImVoxelNet (WACV'2022)\u003C\u002Fa>\u003C\u002Fli>\n          \u003Cli>\u003Ca href=\"configs\u002Fsmoke\">SMOKE (CVPRW'2020)\u003C\u002Fa>\u003C\u002Fli>\n          \u003Cli>\u003Ca href=\"configs\u002Ffcos3d\">FCOS3D (ICCVW'2021)\u003C\u002Fa>\u003C\u002Fli>\n          \u003Cli>\u003Ca href=\"configs\u002Fpgd\">PGD (CoRL'2021)\u003C\u002Fa>\u003C\u002Fli>\n          \u003Cli>\u003Ca href=\"configs\u002Fmonoflex\">MonoFlex (CVPR'2021)\u003C\u002Fa>\u003C\u002Fli>\n          \u003Cli>\u003Ca href=\"projects\u002FDETR3D\">DETR3D (CoRL'2021)\u003C\u002Fa>\u003C\u002Fli>\n          \u003Cli>\u003Ca href=\"projects\u002FPETR\">PETR (ECCV'2022)\u003C\u002Fa>\u003C\u002Fli>\n        \u003C\u002Ful>\n        \u003Cli>\u003Cb>Indoor\u003C\u002Fb>\u003C\u002Fli>\n        \u003Cul>\n          \u003Cli>\u003Ca href=\"configs\u002Fimvoxelnet\">ImVoxelNet (WACV'2022)\u003C\u002Fa>\u003C\u002Fli>\n        \u003C\u002Ful>\n      \u003C\u002Ftd>\n      \u003Ctd>\n        \u003Cli>\u003Cb>Outdoor\u003C\u002Fb>\u003C\u002Fli>\n        \u003Cul>\n          \u003Cli>\u003Ca href=\"configs\u002Fmvxnet\">MVXNet (ICRA'2019)\u003C\u002Fa>\u003C\u002Fli>\n          \u003Cli>\u003Ca href=\"projects\u002FBEVFusion\">BEVFusion (ICRA'2023)\u003C\u002Fa>\u003C\u002Fli>\n        \u003C\u002Ful>\n        \u003Cli>\u003Cb>Indoor\u003C\u002Fb>\u003C\u002Fli>\n        \u003Cul>\n          \u003Cli>\u003Ca href=\"configs\u002Fimvotenet\">ImVoteNet (CVPR'2020)\u003C\u002Fa>\u003C\u002Fli>\n        \u003C\u002Ful>\n      \u003C\u002Ftd>\n      \u003Ctd>\n        \u003Cli>\u003Cb>Outdoor\u003C\u002Fb>\u003C\u002Fli>\n        \u003Cul>\n          \u003Cli>\u003Ca href=\"configs\u002Fminkunet\">MinkUNet (CVPR'2019)\u003C\u002Fa>\u003C\u002Fli>\n          \u003Cli>\u003Ca href=\"configs\u002Fspvcnn\">SPVCNN (ECCV'2020)\u003C\u002Fa>\u003C\u002Fli>\n          \u003Cli>\u003Ca href=\"configs\u002Fcylinder3d\">Cylinder3D (CVPR'2021)\u003C\u002Fa>\u003C\u002Fli>\n          \u003Cli>\u003Ca href=\"projects\u002FTPVFormer\">TPVFormer (CVPR'2023)\u003C\u002Fa>\u003C\u002Fli>\n        \u003C\u002Ful>\n        \u003Cli>\u003Cb>Indoor\u003C\u002Fb>\u003C\u002Fli>\n        \u003Cul>\n          \u003Cli>\u003Ca href=\"configs\u002Fpointnet2\">PointNet++ (NeurIPS'2017)\u003C\u002Fa>\u003C\u002Fli>\n          \u003Cli>\u003Ca href=\"configs\u002Fpaconv\">PAConv (CVPR'2021)\u003C\u002Fa>\u003C\u002Fli>\n          \u003Cli>\u003Ca href=\"configs\u002Fdgcnn\">DGCNN (TOG'2019)\u003C\u002Fa>\u003C\u002Fli>\n        \u003C\u002Ful>\n      \u003C\u002Ful>\n      \u003C\u002Ftd>\n    \u003C\u002Ftr>\n\u003C\u002Ftd>\n    \u003C\u002Ftr>\n  \u003C\u002Ftbody>\n\u003C\u002Ftable>\n\n|               | ResNet | VoVNet | Swin-T | PointNet++ | SECOND | DGCNN | RegNetX | DLA | MinkResNet | Cylinder3D | MinkUNet |\n| :-----------: | :----: | :----: | :----: | :--------: | :----: | :---: | :-----: | :-: | :--------: | :--------: | :------: |\n|    SECOND     |   ✗    |   ✗    |   ✗    |     ✗      |   ✓    |   ✗   |    ✗    |  ✗  |     ✗      |     ✗      |    ✗     |\n| PointPillars  |   ✗    |   ✗    |   ✗    |     ✗      |   ✓    |   ✗   |    ✓    |  ✗  |     ✗      |     ✗      |    ✗     |\n|  FreeAnchor   |   ✗    |   ✗    |   ✗    |     ✗      |   ✗    |   ✗   |    ✓    |  ✗  |     ✗      |     ✗      |    ✗     |\n|    VoteNet    |   ✗    |   ✗    |   ✗    |     ✓      |   ✗    |   ✗   |    ✗    |  ✗  |     ✗      |     ✗      |    ✗     |\n|    H3DNet     |   ✗    |   ✗    |   ✗    |     ✓      |   ✗    |   ✗   |    ✗    |  ✗  |     ✗      |     ✗      |    ✗     |\n|     3DSSD     |   ✗    |   ✗    |   ✗    |     ✓      |   ✗    |   ✗   |    ✗    |  ✗  |     ✗      |     ✗      |    ✗     |\n|    Part-A2    |   ✗    |   ✗    |   ✗    |     ✗      |   ✓    |   ✗   |    ✗    |  ✗  |     ✗      |     ✗      |    ✗     |\n|    MVXNet     |   ✓    |   ✗    |   ✗    |     ✗      |   ✓    |   ✗   |    ✗    |  ✗  |     ✗      |     ✗      |    ✗     |\n|  CenterPoint  |   ✗    |   ✗    |   ✗    |     ✗      |   ✓    |   ✗   |    ✗    |  ✗  |     ✗      |     ✗      |    ✗     |\n|      SSN      |   ✗    |   ✗    |   ✗    |     ✗      |   ✗    |   ✗   |    ✓    |  ✗  |     ✗      |     ✗      |    ✗     |\n|   ImVoteNet   |   ✓    |   ✗    |   ✗    |     ✓      |   ✗    |   ✗   |    ✗    |  ✗  |     ✗      |     ✗      |    ✗     |\n|    FCOS3D     |   ✓    |   ✗    |   ✗    |     ✗      |   ✗    |   ✗   |    ✗    |  ✗  |     ✗      |     ✗      |    ✗     |\n|  PointNet++   |   ✗    |   ✗    |   ✗    |     ✓      |   ✗    |   ✗   |    ✗    |  ✗  |     ✗      |     ✗      |    ✗     |\n| Group-Free-3D |   ✗    |   ✗    |   ✗    |     ✓      |   ✗    |   ✗   |    ✗    |  ✗  |     ✗      |     ✗      |    ✗     |\n|  ImVoxelNet   |   ✓    |   ✗    |   ✗    |     ✗      |   ✗    |   ✗   |    ✗    |  ✗  |     ✗      |     ✗      |    ✗     |\n|    PAConv     |   ✗    |   ✗    |   ✗    |     ✓      |   ✗    |   ✗   |    ✗    |  ✗  |     ✗      |     ✗      |    ✗     |\n|     DGCNN     |   ✗    |   ✗    |   ✗    |     ✗      |   ✗    |   ✓   |    ✗    |  ✗  |     ✗      |     ✗      |    ✗     |\n|     SMOKE     |   ✗    |   ✗    |   ✗    |     ✗      |   ✗    |   ✗   |    ✗    |  ✓  |     ✗      |     ✗      |    ✗     |\n|      PGD      |   ✓    |   ✗    |   ✗    |     ✗      |   ✗    |   ✗   |    ✗    |  ✗  |     ✗      |     ✗      |    ✗     |\n|   MonoFlex    |   ✗    |   ✗    |   ✗    |     ✗      |   ✗    |   ✗   |    ✗    |  ✓  |     ✗      |     ✗      |    ✗     |\n|    SA-SSD     |   ✗    |   ✗    |   ✗    |     ✗      |   ✓    |   ✗   |    ✗    |  ✗  |     ✗      |     ✗      |    ✗     |\n|    FCAF3D     |   ✗    |   ✗    |   ✗    |     ✗      |   ✗    |   ✗   |    ✗    |  ✗  |     ✓      |     ✗      |    ✗     |\n|    PV-RCNN    |   ✗    |   ✗    |   ✗    |     ✗      |   ✓    |   ✗   |    ✗    |  ✗  |     ✗      |     ✗      |    ✗     |\n|  Cylinder3D   |   ✗    |   ✗    |   ✗    |     ✗      |   ✗    |   ✗   |    ✗    |  ✗  |     ✗      |     ✓      |    ✗     |\n|   MinkUNet    |   ✗    |   ✗    |   ✗    |     ✗      |   ✗    |   ✗   |    ✗    |  ✗  |     ✗      |     ✗      |    ✓     |\n|    SPVCNN     |   ✗    |   ✗    |   ✗    |     ✗      |   ✗    |   ✗   |    ✗    |  ✗  |     ✗      |     ✗      |    ✓     |\n|   BEVFusion   |   ✗    |   ✗    |   ✓    |     ✗      |   ✓    |   ✗   |    ✗    |  ✗  |     ✗      |     ✗      |    ✗     |\n| CenterFormer  |   ✗    |   ✗    |   ✗    |     ✗      |   ✓    |   ✗   |    ✗    |  ✗  |     ✗      |     ✗      |    ✗     |\n|     TR3D      |   ✗    |   ✗    |   ✗    |     ✗      |   ✗    |   ✗   |    ✗    |  ✗  |     ✓      |     ✗      |    ✗     |\n|    DETR3D     |   ✓    |   ✓    |   ✗    |     ✗      |   ✗    |   ✗   |    ✗    |  ✗  |     ✗      |     ✗      |    ✗     |\n|     PETR      |   ✗    |   ✓    |   ✗    |     ✗      |   ✗    |   ✗   |    ✗    |  ✗  |     ✗      |     ✗      |    ✗     |\n|   TPVFormer   |   ✓    |   ✗    |   ✗    |     ✗      |   ✗    |   ✗   |    ✗    |  ✗  |     ✗      |     ✗      |    ✗     |\n\n**Note:** All the about **500+ models, methods of 90+ papers** in 2D detection supported by [MMDetection](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection\u002Fblob\u002F3.x\u002Fdocs\u002Fen\u002Fmodel_zoo.md) can be trained or used in this codebase.\n\n## FAQ\n\nPlease refer to [FAQ](docs\u002Fen\u002Fnotes\u002Ffaq.md) for frequently asked questions.\n\n## Contributing\n\nWe appreciate all contributions to improve MMDetection3D. Please refer to [CONTRIBUTING.md](docs\u002Fen\u002Fnotes\u002Fcontribution_guides.md) for the contributing guideline.\n\n## Acknowledgement\n\nMMDetection3D is an open source project that is contributed by researchers and engineers from various colleges and companies. We appreciate all the contributors as well as users who give valuable feedbacks. We wish that the toolbox and benchmark could serve the growing research community by providing a flexible toolkit to reimplement existing methods and develop their own new 3D detectors.\n\n## Citation\n\nIf you find this project useful in your research, please consider cite:\n\n```latex\n@misc{mmdet3d2020,\n    title={{MMDetection3D: OpenMMLab} next-generation platform for general {3D} object detection},\n    author={MMDetection3D Contributors},\n    howpublished = {\\url{https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d}},\n    year={2020}\n}\n```\n\n## License\n\nThis project is released under the [Apache 2.0 license](LICENSE).\n\n## Projects in OpenMMLab\n\n- [MMEngine](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmengine): OpenMMLab foundational library for training deep learning models.\n- [MMCV](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmcv): OpenMMLab foundational library for computer vision.\n- [MMEval](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmeval): A unified evaluation library for multiple machine learning libraries.\n- [MIM](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmim): MIM installs OpenMMLab packages.\n- [MMPreTrain](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmpretrain): OpenMMLab pre-training toolbox and benchmark.\n- [MMDetection](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection): OpenMMLab detection toolbox and benchmark.\n- [MMDetection3D](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d): OpenMMLab's next-generation platform for general 3D object detection.\n- [MMRotate](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmrotate): OpenMMLab rotated object detection toolbox and benchmark.\n- [MMYOLO](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmyolo): OpenMMLab YOLO series toolbox and benchmark.\n- [MMSegmentation](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmsegmentation): OpenMMLab semantic segmentation toolbox and benchmark.\n- [MMOCR](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmocr): OpenMMLab text detection, recognition, and understanding toolbox.\n- [MMPose](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmpose): OpenMMLab pose estimation toolbox and benchmark.\n- [MMHuman3D](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmhuman3d): OpenMMLab 3D human parametric model toolbox and benchmark.\n- [MMSelfSup](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmselfsup): OpenMMLab self-supervised learning toolbox and benchmark.\n- [MMRazor](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmrazor): OpenMMLab model compression toolbox and benchmark.\n- [MMFewShot](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmfewshot): OpenMMLab fewshot learning toolbox and benchmark.\n- [MMAction2](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmaction2): OpenMMLab's next-generation action understanding toolbox and benchmark.\n- [MMTracking](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmtracking): OpenMMLab video perception toolbox and benchmark.\n- [MMFlow](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmflow): OpenMMLab optical flow toolbox and benchmark.\n- [MMagic](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmagic): Open**MM**Lab **A**dvanced, **G**enerative and **I**ntelligent **C**reation toolbox.\n- [MMGeneration](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmgeneration): OpenMMLab image and video generative models toolbox.\n- [MMDeploy](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdeploy): OpenMMLab model deployment framework.\n","\u003Cdiv align=\"center\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fopen-mmlab_mmdetection3d_readme_bc52b0de66dc.png\" width=\"600\"\u002F>\n  \u003Cdiv> \u003C\u002Fdiv>\n  \u003Cdiv align=\"center\">\n    \u003Cb>\u003Cfont size=\"5\">OpenMMLab 官网\u003C\u002Ffont>\u003C\u002Fb>\n    \u003Csup>\n      \u003Ca href=\"https:\u002F\u002Fopenmmlab.com\">\n        \u003Ci>\u003Cfont size=\"4\">热门\u003C\u002Ffont>\u003C\u002Fi>\n      \u003C\u002Fa>\n    \u003C\u002Fsup>\n        \n    \u003Cb>\u003Cfont size=\"5\">OpenMMLab 平台\u003C\u002Ffont>\u003C\u002Fb>\n    \u003Csup>\n      \u003Ca href=\"https:\u002F\u002Fplatform.openmmlab.com\">\n        \u003Ci>\u003Cfont size=\"4\">立即体验\u003C\u002Ffont>\u003C\u002Fi>\n      \u003C\u002Fa>\n    \u003C\u002Fsup>\n  \u003C\u002Fdiv>\n  \u003Cdiv> \u003C\u002Fdiv>\n\n[![PyPI](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fv\u002Fmmdet3d)](https:\u002F\u002Fpypi.org\u002Fproject\u002Fmmdet3d)\n[![docs](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdocs-latest-blue)](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002F)\n[![badge](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fworkflows\u002Fbuild\u002Fbadge.svg)](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Factions)\n[![codecov](https:\u002F\u002Fcodecov.io\u002Fgh\u002Fopen-mmlab\u002Fmmdetection3d\u002Fbranch\u002Fmain\u002Fgraph\u002Fbadge.svg)](https:\u002F\u002Fcodecov.io\u002Fgh\u002Fopen-mmlab\u002Fmmdetection3d)\n[![license](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Flicense\u002Fopen-mmlab\u002Fmmdetection3d.svg)](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fblob\u002Fmain\u002FLICENSE)\n[![open issues](https:\u002F\u002Fisitmaintained.com\u002Fbadge\u002Fopen\u002Fopen-mmlab\u002Fmmdetection3d.svg)](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fissues)\n[![issue resolution](https:\u002F\u002Fisitmaintained.com\u002Fbadge\u002Fresolution\u002Fopen-mmlab\u002Fmmdetection3d.svg)](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fissues)\n\n[📘文档](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002F) |\n[🛠️安装](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fget_started.html) |\n[👀模型库](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fmodel_zoo.html) |\n[🆕更新消息](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fnotes\u002Fchangelog.html) |\n[🚀正在进行的项目](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fprojects) |\n[🤔报告问题](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fissues\u002Fnew\u002Fchoose)\n\n\u003C\u002Fdiv>\n\n\u003Cdiv align=\"center\">\n\nEnglish | [简体中文](README_zh-CN.md)\n\n\u003C\u002Fdiv>\n\n\u003Cdiv align=\"center\">\n  \u003Ca href=\"https:\u002F\u002Fopenmmlab.medium.com\u002F\" style=\"text-decoration:none;\">\n    \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fopen-mmlab_mmdetection3d_readme_062337b0e5ec.png\" width=\"3%\" alt=\"\" \u002F>\u003C\u002Fa>\n  \u003Cimg src=\"https:\u002F\u002Fuser-images.githubusercontent.com\u002F25839884\u002F218346358-56cc8e2f-a2b8-487f-9088-32480cceabcf.png\" width=\"3%\" alt=\"\" \u002F>\n  \u003Ca href=\"https:\u002F\u002Fdiscord.com\u002Fchannels\u002F1037617289144569886\u002F1046608014234370059\" style=\"text-decoration:none;\">\n    \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fopen-mmlab_mmdetection3d_readme_6342e5371027.png\" width=\"3%\" alt=\"\" \u002F>\u003C\u002Fa>\n  \u003Cimg src=\"https:\u002F\u002Fuser-images.githubusercontent.com\u002F25839884\u002F218346358-56cc8e2f-a2b8-487f-9088-32480cceabcf.png\" width=\"3%\" alt=\"\" \u002F>\n  \u003Ca href=\"https:\u002F\u002Ftwitter.com\u002FOpenMMLab\" style=\"text-decoration:none;\">\n    \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fopen-mmlab_mmdetection3d_readme_04c3beda0b07.png\" width=\"3%\" alt=\"\" \u002F>\u003C\u002Fa>\n  \u003Cimg src=\"https:\u002F\u002Fuser-images.githubusercontent.com\u002F25839884\u002F218346358-56cc8e2f-a2b8-487f-9088-32480cceabcf.png\" width=\"3%\" alt=\"\" \u002F>\n  \u003Ca href=\"https:\u002F\u002Fwww.youtube.com\u002Fopenmmlab\" style=\"text-decoration:none;\">\n    \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fopen-mmlab_mmdetection3d_readme_204fe79b5a90.png\" width=\"3%\" alt=\"\" \u002F>\u003C\u002Fa>\n  \u003Cimg src=\"https:\u002F\u002Fuser-images.githubusercontent.com\u002F25839884\u002F218346358-56cc8e2f-a2b8-487f-9088-32480cceabcf.png\" width=\"3%\" alt=\"\" \u002F>\n  \u003Ca href=\"https:\u002F\u002Fspace.bilibili.com\u002F1293512903\" style=\"text-decoration:none;\">\n    \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fopen-mmlab_mmdetection3d_readme_8655b6233577.png\" width=\"3%\" alt=\"\" \u002F>\u003C\u002Fa>\n  \u003Cimg src=\"https:\u002F\u002Fuser-images.githubusercontent.com\u002F25839884\u002F218346358-56cc8e2f-a2b8-487f-9088-32480cceabcf.png\" width=\"3%\" alt=\"\" \u002F>\n  \u003Ca href=\"https:\u002F\u002Fwww.zhihu.com\u002Fpeople\u002Fopenmmlab\" style=\"text-decoration:none;\">\n    \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fopen-mmlab_mmdetection3d_readme_447c4737c11f.png\" width=\"3%\" alt=\"\" \u002F>\u003C\u002Fa>\n\u003C\u002Fdiv>\n\n## 简介\n\nMMDetection3D 是一个基于 PyTorch 的开源目标检测工具箱，旨在打造下一代通用 3D 检测平台。它是 [OpenMMLab](https:\u002F\u002Fopenmmlab.com\u002F) 项目的一部分。\n\n主分支支持 **PyTorch 1.8+**。\n\n![demo image](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fopen-mmlab_mmdetection3d_readme_5e593c5a783a.gif)\n\n\u003Cdetails open>\n\u003Csummary>主要特性\u003C\u002Fsummary>\n\n- **开箱即用的多模态\u002F单模态检测器支持**\n\n  直接支持包括 MVXNet、VoteNet、PointPillars 等在内的多模态和单模态检测器。\n\n- **开箱即用的室内\u002F室外 3D 检测支持**\n\n  直接支持流行的室内和室外 3D 检测数据集，包括 ScanNet、SUNRGB-D、Waymo、nuScenes、Lyft 和 KITTI。对于 nuScenes 数据集，我们也支持 [nuImages 数据集](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Ftree\u002Fmain\u002Fconfigs\u002Fnuimages)。\n\n- **与 2D 检测的自然集成**\n\n  [MMDetection](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection\u002Fblob\u002F3.x\u002Fdocs\u002Fen\u002Fmodel_zoo.md) 中支持的 **40 多篇论文中的 300 多种模型和方法**，以及所有模块，都可以在此代码库中进行训练或使用。\n\n- **高效性**\n\n  它的训练速度比其他代码库更快。主要结果如下所示。详细信息请参阅 [benchmark.md](.\u002Fdocs\u002Fen\u002Fnotes\u002Fbenchmarks.md)。我们比较每秒训练的样本数（数值越高越好）。其他代码库不支持的模型以 `✗` 标示。\n\n  |       方法       | MMDetection3D | [OpenPCDet](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002FOpenPCDet) | [votenet](https:\u002F\u002Fgithub.com\u002Ffacebookresearch\u002Fvotenet) | [Det3D](https:\u002F\u002Fgithub.com\u002Fpoodarchu\u002FDet3D) |\n  | :-----------------: | :-----------: | :--------------------------------------------------: | :----------------------------------------------------: | :-----------------------------------------: |\n  |       VoteNet       |      358      |                          ✗                           |                           77                           |                      ✗                      |\n  |  PointPillars-car   |      141      |                          ✗                           |                           ✗                            |                     140                     |\n  | PointPillars-3class |      107      |                          44                          |                           ✗                            |                      ✗                      |\n  |       SECOND        |      40       |                          30                          |                           ✗                            |                      ✗                      |\n  |       Part-A2       |      17       |                          14                          |                           ✗                            |                      ✗                      |\n\n与 [MMDetection](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection) 和 [MMCV](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmcv) 类似，MMDetection3D 也可以作为库来支持在其基础上构建的各种项目。\n\n## 最新动态\n\n### 亮点\n\n在 1.4 版本中，MMDetection3D 对 Waymo 数据集进行了重构，并加速了 Waymo 数据集的预处理、训练\u002F测试设置以及评估流程。此外，我们还扩展了对基于摄像头的 3D 目标检测模型的支持，例如单目和 BEV 模式下的 Waymo 数据集。关于 Waymo 数据的详细说明请参见 [这里](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fadvanced_guides\u002Fdatasets\u002Fwaymo.html)。\n\n另外，在 1.4 版本中，MMDetection3D 提供了 [Waymo-mini](https:\u002F\u002Fdownload.openmmlab.com\u002Fmmdetection3d\u002Fdata\u002Fwaymo_mmdet3d_after_1x4\u002Fwaymo_mini.tar.gz)，帮助社区用户快速上手 Waymo 数据，并用于快速迭代开发。\n\n**v1.4.0** 于 2024 年 8 月 1 日发布：\n\n- 在 `projects` 中支持 [DSVT](\u003C(https:\u002F\u002Farxiv.org\u002Fabs\u002F2301.06051)>) 的训练\n- 在 `projects` 中支持 [Nerf-Det](https:\u002F\u002Farxiv.org\u002Fabs\u002F2307.14620)\n- 重构 Waymo 数据集\n\n**v1.3.0** 于 2023 年 10 月 18 日发布：\n\n- 在 `projects` 中支持 [CENet](https:\u002F\u002Farxiv.org\u002Fabs\u002F2207.12691)\n- 使用新的 3D 推理器增强演示效果\n\n**v1.2.0** 于 2023 年 7 月 4 日发布：\n\n- 在 `mmdet3d\u002Fconfigs` 中支持 [新型配置文件格式](https:\u002F\u002Fmmengine.readthedocs.io\u002Fen\u002Flatest\u002Fadvanced_tutorials\u002Fconfig.html#a-pure-python-style-configuration-file-beta)\n- 在 `projects` 中支持 [DSVT](\u003C(https:\u002F\u002Farxiv.org\u002Fabs\u002F2301.06051)>) 的推理\n- 支持使用 `mim` 从 [OpenDataLab](https:\u002F\u002Fopendatalab.com\u002F) 下载数据集\n\n**v1.1.1** 于 2023 年 5 月 30 日发布：\n\n- 在 `projects` 中支持 [TPVFormer](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2302.07817.pdf)\n- 在 `projects` 中支持 BEVFusion 的训练\n- 支持基于激光雷达的 3D 语义分割基准测试\n\n## 安装\n\n请参阅 [安装指南](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fget_started.html) 获取安装说明。\n\n## 入门指南\n\n有关详细的用户指南和高级指南，请参阅我们的[文档](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002F)：\n\n\u003Cdetails>\n\u003Csummary>用户指南\u003C\u002Fsummary>\n\n- [训练与测试](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fuser_guides\u002Findex.html#train-test)\n  - [了解配置文件](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fuser_guides\u002Fconfig.html)\n  - [坐标系](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fuser_guides\u002Fcoord_sys_tutorial.html)\n  - [数据集准备](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fuser_guides\u002Fdataset_prepare.html)\n  - [自定义数据流水线](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fuser_guides\u002Fdata_pipeline.html)\n  - [在标准数据集上进行训练和测试](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fuser_guides\u002Ftrain_test.html)\n  - [推理](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fuser_guides\u002Finference.html)\n  - [使用自定义数据集进行训练](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fuser_guides\u002Fnew_data_model.html)\n- [实用工具](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fuser_guides\u002Findex.html#useful-tools)\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>高级指南\u003C\u002Fsummary>\n\n- [数据集](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fadvanced_guides\u002Findex.html#datasets)\n  - [KITTI 数据集](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fadvanced_guides\u002Fdatasets\u002Fkitti.html)\n  - [NuScenes 数据集](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fadvanced_guides\u002Fdatasets\u002Fnuscenes.html)\n  - [Lyft 数据集](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fadvanced_guides\u002Fdatasets\u002Flyft.html)\n  - [Waymo 数据集](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fadvanced_guides\u002Fdatasets\u002Fwaymo.html)\n  - [SUN RGB-D 数据集](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fadvanced_guides\u002Fdatasets\u002Fsunrgbd.html)\n  - [ScanNet 数据集](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fadvanced_guides\u002Fdatasets\u002Fscannet.html)\n  - [S3DIS 数据集](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fadvanced_guides\u002Fdatasets\u002Fs3dis.html)\n  - [SemanticKITTI 数据集](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fadvanced_guides\u002Fdatasets\u002Fsemantickitti.html)\n- [支持的任务](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fadvanced_guides\u002Findex.html#supported-tasks)\n  - [基于 LiDAR 的 3D 检测](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fadvanced_guides\u002Fsupported_tasks\u002Flidar_det3d.html)\n  - [基于视觉的 3D 检测](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fadvanced_guides\u002Fsupported_tasks\u002Fvision_det3d.html)\n  - [基于 LiDAR 的 3D 语义分割](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fadvanced_guides\u002Fsupported_tasks\u002Flidar_sem_seg3d.html)\n- [定制化](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fadvanced_guides\u002Findex.html#customization)\n  - [自定义数据集](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fadvanced_guides\u002Fcustomize_dataset.html)\n  - [自定义模型](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fadvanced_guides\u002Fcustomize_models.html)\n  - [自定义运行时设置](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002Flatest\u002Fadvanced_guides\u002Fcustomize_runtime.html)\n\n\u003C\u002Fdetails>\n\n## 基准与模型库概览\n\n结果和模型可在[模型库](docs\u002Fen\u002Fmodel_zoo.md)中找到。\n\n\u003Cdiv align=\"center\">\n  \u003Cb>组件\u003C\u002Fb>\n\u003C\u002Fdiv>\n\u003Ctable align=\"center\">\n  \u003Ctbody>\n    \u003Ctr align=\"center\" valign=\"bottom\">\n      \u003Ctd>\n        \u003Cb>骨干网络\u003C\u002Fb>\n      \u003C\u002Ftd>\n      \u003Ctd>\n        \u003Cb>检测头\u003C\u002Fb>\n      \u003C\u002Ftd>\n      \u003Ctd>\n        \u003Cb>特征提取模块\u003C\u002Fb>\n      \u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr valign=\"top\">\n      \u003Ctd>\n      \u003Cul>\n        \u003Cli>\u003Ca href=\"configs\u002Fpointnet2\">PointNet (CVPR'2017)\u003C\u002Fa>\u003C\u002Fli>\n        \u003Cli>\u003Ca href=\"configs\u002Fpointnet2\">PointNet++ (NeurIPS'2017)\u003C\u002Fa>\u003C\u002Fli>\n        \u003Cli>\u003Ca href=\"configs\u002Fregnet\">RegNet (CVPR'2020)\u003C\u002Fa>\u003C\u002Fli>\n        \u003Cli>\u003Ca href=\"configs\u002Fdgcnn\">DGCNN (TOG'2019)\u003C\u002Fa>\u003C\u002Fli>\n        \u003Cli>DLA (CVPR'2018)\u003C\u002Fli>\n        \u003Cli>MinkResNet (CVPR'2019)\u003C\u002Fli>\n        \u003Cli>\u003Ca href=\"configs\u002Fminkunet\">MinkUNet (CVPR'2019)\u003C\u002Fa>\u003C\u002Fli>\n        \u003Cli>\u003Ca href=\"configs\u002Fcylinder3d\">Cylinder3D (CVPR'2021)\u003C\u002Fa>\u003C\u002Fli>\n      \u003C\u002Ful>\n      \u003C\u002Ftd>\n      \u003Ctd>\n      \u003Cul>\n        \u003Cli>\u003Ca href=\"configs\u002Ffree_anchor\">FreeAnchor (NeurIPS'2019)\u003C\u002Fa>\u003C\u002Fli>\n      \u003C\u002Ful>\n      \u003C\u002Ftd>\n      \u003Ctd>\n      \u003Cul>\n        \u003Cli>\u003Ca href=\"configs\u002Fdynamic_voxelization\">动态体素化 (CoRL'2019)\u003C\u002Fa>\u003C\u002Fli>\n      \u003C\u002Ful>\n      \u003C\u002Ftd>\n    \u003C\u002Ftr>\n\u003C\u002Ftd>\n    \u003C\u002Ftr>\n  \u003C\u002Ftbody>\n\u003C\u002Ftable>\n\n\u003Cdiv align=\"center\">\n  \u003Cb>架构\u003C\u002Fb>\n\u003C\u002Fdiv>\n\u003Ctable align=\"center\">\n  \u003Ctbody>\n    \u003Ctr align=\"center\" valign=\"middle\">\n      \u003Ctd>\n        \u003Cb>基于LiDAR的3D目标检测\u003C\u002Fb>\n      \u003C\u002Ftd>\n      \u003Ctd>\n        \u003Cb>基于相机的3D目标检测\u003C\u002Fb>\n      \u003C\u002Ftd>\n      \u003Ctd>\n        \u003Cb>多模态3D目标检测\u003C\u002Fb>\n      \u003C\u002Ftd>\n      \u003Ctd>\n        \u003Cb>3D语义分割\u003C\u002Fb>\n      \u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr valign=\"top\">\n      \u003Ctd>\n        \u003Cli>\u003Cb>室外\u003C\u002Fb>\u003C\u002Fli>\n        \u003Cul>\n            \u003Cli>\u003Ca href=\"configs\u002Fsecond\">SECOND (Sensor'2018)\u003C\u002Fa>\u003C\u002Fli>\n            \u003Cli>\u003Ca href=\"configs\u002Fpointpillars\">PointPillars (CVPR'2019)\u003C\u002Fa>\u003C\u002Fli>\n            \u003Cli>\u003Ca href=\"configs\u002Fssn\">SSN (ECCV'2020)\u003C\u002Fa>\u003C\u002Fli>\n            \u003Cli>\u003Ca href=\"configs\u002F3dssd\">3DSSD (CVPR'2020)\u003C\u002Fa>\u003C\u002Fli>\n            \u003Cli>\u003Ca href=\"configs\u002Fsassd\">SA-SSD (CVPR'2020)\u003C\u002Fa>\u003C\u002Fli>\n            \u003Cli>\u003Ca href=\"configs\u002Fpoint_rcnn\">PointRCNN (CVPR'2019)\u003C\u002Fa>\u003C\u002Fli>\n            \u003Cli>\u003Ca href=\"configs\u002Fparta2\">Part-A2 (TPAMI'2020)\u003C\u002Fa>\u003C\u002Fli>\n            \u003Cli>\u003Ca href=\"configs\u002Fcenterpoint\">CenterPoint (CVPR'2021)\u003C\u002Fa>\u003C\u002Fli>\n            \u003Cli>\u003Ca href=\"configs\u002Fpv_rcnn\">PV-RCNN (CVPR'2020)\u003C\u002Fa>\u003C\u002Fli>\n            \u003Cli>\u003Ca href=\"projects\u002FCenterFormer\">CenterFormer (ECCV'2022)\u003C\u002Fa>\u003C\u002Fli>\n        \u003C\u002Ful>\n        \u003Cli>\u003Cb>室内\u003C\u002Fb>\u003C\u002Fli>\n        \u003Cul>\n            \u003Cli>\u003Ca href=\"configs\u002Fvotenet\">VoteNet (ICCV'2019)\u003C\u002Fa>\u003C\u002Fli>\n            \u003Cli>\u003Ca href=\"configs\u002Fh3dnet\">H3DNet (ECCV'2020)\u003C\u002Fa>\u003C\u002Fli>\n            \u003Cli>\u003Ca href=\"configs\u002Fgroupfree3d\">Group-Free-3D (ICCV'2021)\u003C\u002Fa>\u003C\u002Fli>\n            \u003Cli>\u003Ca href=\"configs\u002Ffcaf3d\">FCAF3D (ECCV'2022)\u003C\u002Fa>\u003C\u002Fli>\n            \u003Cli>\u003Ca href=\"projects\u002FTR3D\">TR3D (ArXiv'2023)\u003C\u002Fa>\u003C\u002Fli>\n      \u003C\u002Ful>\n      \u003C\u002Ftd>\n      \u003Ctd>\n        \u003Cli>\u003Cb>室外\u003C\u002Fb>\u003C\u002Fli>\n        \u003Cul>\n          \u003Cli>\u003Ca href=\"configs\u002Fimvoxelnet\">ImVoxelNet (WACV'2022)\u003C\u002Fa>\u003C\u002Fli>\n          \u003Cli>\u003Ca href=\"configs\u002Fsmoke\">SMOKE (CVPRW'2020)\u003C\u002Fa>\u003C\u002Fli>\n          \u003Cli>\u003Ca href=\"configs\u002Ffcos3d\">FCOS3D (ICCVW'2021)\u003C\u002Fa>\u003C\u002Fli>\n          \u003Cli>\u003Ca href=\"configs\u002Fpgd\">PGD (CoRL'2021)\u003C\u002Fa>\u003C\u002Fli>\n          \u003Cli>\u003Ca href=\"configs\u002Fmonoflex\">MonoFlex (CVPR'2021)\u003C\u002Fa>\u003C\u002Fli>\n          \u003Cli>\u003Ca href=\"projects\u002FDETR3D\">DETR3D (CoRL'2021)\u003C\u002Fa>\u003C\u002Fli>\n          \u003Cli>\u003Ca href=\"projects\u002FPETR\">PETR (ECCV'2022)\u003C\u002Fa>\u003C\u002Fli>\n        \u003C\u002Ful>\n        \u003Cli>\u003Cb>室内\u003C\u002Fb>\u003C\u002Fli>\n        \u003Cul>\n          \u003Cli>\u003Ca href=\"configs\u002Fimvoxelnet\">ImVoxelNet (WACV'2022)\u003C\u002Fa>\u003C\u002Fli>\n        \u003C\u002Ful>\n      \u003C\u002Ftd>\n      \u003Ctd>\n        \u003Cli>\u003Cb>室外\u003C\u002Fb>\u003C\u002Fli>\n        \u003Cul>\n          \u003Cli>\u003Ca href=\"configs\u002Fmvxnet\">MVXNet (ICRA'2019)\u003C\u002Fa>\u003C\u002Fli>\n          \u003Cli>\u003Ca href=\"projects\u002FBEVFusion\">BEVFusion (ICRA'2023)\u003C\u002Fa>\u003C\u002Fli>\n        \u003C\u002Ful>\n        \u003Cli>\u003Cb>室内\u003C\u002Fb>\u003C\u002Fli>\n        \u003Cul>\n          \u003Cli>\u003Ca href=\"configs\u002Fimvotenet\">ImVoteNet (CVPR'2020)\u003C\u002Fa>\u003C\u002Fli>\n        \u003C\u002Ful>\n      \u003C\u002Ftd>\n      \u003Ctd>\n        \u003Cli>\u003Cb>室外\u003C\u002Fb>\u003C\u002Fli>\n        \u003Cul>\n          \u003Cli>\u003Ca href=\"configs\u002Fminkunet\">MinkUNet (CVPR'2019)\u003C\u002Fa>\u003C\u002Fli>\n          \u003Cli>\u003Ca href=\"configs\u002Fspvcnn\">SPVCNN (ECCV'2020)\u003C\u002Fa>\u003C\u002Fli>\n          \u003Cli>\u003Ca href=\"configs\u002Fcylinder3d\">Cylinder3D (CVPR'2021)\u003C\u002Fa>\u003C\u002Fli>\n          \u003Cli>\u003Ca href=\"projects\u002FTPVFormer\">TPVFormer (CVPR'2023)\u003C\u002Fa>\u003C\u002Fli>\n        \u003C\u002Ful>\n        \u003Cli>\u003Cb>室内\u003C\u002Fb>\u003C\u002Fli>\n        \u003Cul>\n          \u003Cli>\u003Ca href=\"configs\u002Fpointnet2\">PointNet++ (NeurIPS'2017)\u003C\u002Fa>\u003C\u002Fli>\n          \u003Cli>\u003Ca href=\"configs\u002Fpaconv\">PAConv (CVPR'2021)\u003C\u002Fa>\u003C\u002Fli>\n          \u003Cli>\u003Ca href=\"configs\u002Fdgcnn\">DGCNN (TOG'2019)\u003C\u002Fa>\u003C\u002Fli>\n        \u003C\u002Ful>\n      \u003C\u002Ftd>\n    \u003C\u002Ftr>\n\u003C\u002Ftd>\n    \u003C\u002Ftr>\n  \u003C\u002Ftbody>\n\u003C\u002Ftable>\n\n|               | ResNet | VoVNet | Swin-T | PointNet++ | SECOND | DGCNN | RegNetX | DLA | MinkResNet | Cylinder3D | MinkUNet |\n| :-----------: | :----: | :----: | :----: | :--------: | :----: | :---: | :-----: | :-: | :--------: | :--------: | :------: |\n|    SECOND     |   ✗    |   ✗    |   ✗    |     ✗      |   ✓    |   ✗   |    ✗    |  ✗  |     ✗      |     ✗      |    ✗     |\n| PointPillars  |   ✗    |   ✗    |   ✗    |     ✗      |   ✓    |   ✗   |    ✓    |  ✗  |     ✗      |     ✗      |    ✗     |\n|  FreeAnchor   |   ✗    |   ✗    |   ✗    |     ✗      |   ✗    |   ✗   |    ✓    |  ✗  |     ✗      |     ✗      |    ✗     |\n|    VoteNet    |   ✗    |   ✗    |   ✗    |     ✓      |   ✗    |   ✗   |    ✗    |  ✗  |     ✗      |     ✗      |    ✗     |\n|    H3DNet     |   ✗    |   ✗    |   ✗    |     ✓      |   ✗    |   ✗   |    ✗    |  ✗  |     ✗      |     ✗      |    ✗     |\n|     3DSSD     |   ✗    |   ✗    |   ✗    |     ✓      |   ✗    |   ✗   |    ✗    |  ✗  |     ✗      |     ✗      |    ✗     |\n|    Part-A2    |   ✗    |   ✗    |   ✗    |     ✗      |   ✓    |   ✗   |    ✗    |  ✗  |     ✗      |     ✗      |    ✗     |\n|    MVXNet     |   ✓    |   ✗    |   ✗    |     ✗      |   ✓    |   ✗   |    ✗    |  ✗  |     ✗      |     ✗      |    ✗     |\n|  CenterPoint  |   ✗    |   ✗    |   ✗    |     ✗      |   ✓    |   ✗   |    ✗    |  ✗  |     ✗      |     ✗      |    ✗     |\n|      SSN      |   ✗    |   ✗    |   ✗    |     ✗      |   ✗    |   ✗   |    ✓    |  ✗  |     ✗      |     ✗      |    ✗     |\n|   ImVoteNet   |   ✓    |   ✗    |   ✗    |     ✓      |   ✗    |   ✗   |    ✗    |  ✗  |     ✗      |     ✗      |    ✗     |\n|    FCOS3D     |   ✓    |   ✗    |   ✗    |     ✗      |   ✗    |   ✗   |    ✗    |  ✗  |     ✗      |     ✗      |    ✗     |\n|  PointNet++   |   ✗    |   ✗    |   ✗    |     ✓      |   ✗    |   ✗   |    ✗    |  ✗  |     ✗      |     ✗      |    ✗     |\n| Group-Free-3D |   ✗    |   ✗    |   ✗    |     ✓      |   ✗    |   ✗   |    ✗    |  ✗  |     ✗      |     ✗      |    ✗     |\n|  ImVoxelNet   |   ✓    |   ✗    |   ✗    |     ✗      |   ✗    |   ✗   |    ✗    |  ✗  |     ✗      |     ✗      |    ✗     |\n|    PAConv     |   ✗    |   ✗    |   ✗    |     ✓      |   ✗    |   ✗   |    ✗    |  ✗  |     ✗      |     ✗      |    ✗     |\n|     DGCNN     |   ✗    |   ✗    |   ✗    |     ✗      |   ✗    |   ✓   |    ✗    |  ✗  |     ✗      |     ✗      |    ✗     |\n|     SMOKE     |   ✗    |   ✗    |   ✗    |     ✗      |   ✗    |   ✗   |    ✗    |  ✓  |     ✗      |     ✗      |    ✗     |\n|      PGD      |   ✓    |   ✗    |   ✗    |     ✗      |   ✗    |   ✗   |    ✗    |  ✗  |     ✗      |     ✗      |    ✗     |\n|   MonoFlex    |   ✗    |   ✗    |   ✗    |     ✗      |   ✗    |   ✗   |    ✗    |  ✓  |     ✗      |     ✗      |    ✗     |\n|    SA-SSD     |   ✗    |   ✗    |   ✗    |     ✗      |   ✓    |   ✗   |    ✗    |  ✗  |     ✗      |     ✗      |    ✗     |\n|    FCAF3D     |   ✗    |   ✗    |   ✗    |     ✗      |   ✗    |   ✗   |    ✗    |  ✗  |     ✓      |     ✗      |    ✗     |\n|    PV-RCNN    |   ✗    |   ✗    |   ✗    |     ✗      |   ✓    |   ✗   |    ✗    |  ✗  |     ✗      |     ✗      |    ✗     |\n|  Cylinder3D   |   ✗    |   ✗    |   ✗    |     ✗      |   ✗    |   ✗   |    ✗    |  ✗  |     ✗      |     ✓      |    ✗     |\n|   MinkUNet    |   ✗    |   ✗    |   ✗    |     ✗      |   ✗    |   ✗   |    ✗    |  ✗  |     ✗      |     ✗      |    ✓     |\n|    SPVCNN     |   ✗    |   ✗    |   ✗    |     ✗      |   ✗    |   ✗   |    ✗    |  ✗  |     ✗      |     ✗      |    ✓     |\n|   BEVFusion   |   ✗    |   ✗    |   ✓    |     ✗      |   ✓    |   ✗   |    ✗    |  ✗  |     ✗      |     ✗      |    ✗     |\n| CenterFormer  |   ✗    |   ✗    |   ✗    |     ✗      |   ✓    |   ✗   |    ✗    |  ✗  |     ✗      |     ✗      |    ✗     |\n|     TR3D      |   ✗    |   ✗    |   ✗    |     ✗      |   ✗    |   ✗   |    ✗    |  ✗  |     ✓      |     ✗      |    ✗     |\n|    DETR3D     |   ✓    |   ✓    |   ✗    |     ✗      |   ✗    |   ✗   |    ✗    |  ✗  |     ✗      |     ✗      |    ✗     |\n|     PETR      |   ✗    |   ✓    |   ✗    |     ✗      |   ✗    |   ✗   |    ✗    |  ✗  |     ✗      |     ✗      |    ✗     |\n|   TPVFormer   |   ✓    |   ✗    |   ✗    |     ✗      |   ✗    |   ✗   |    ✗    |  ✗  |     ✗      |     ✗      |    ✗     |\n\n**注：** [MMDetection](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection\u002Fblob\u002F3.x\u002Fdocs\u002Fen\u002Fmodel_zoo.md) 支持的 **500 多种模型和 90 多篇论文中的方法**，均可在此代码库中进行训练或使用。\n\n\n\n## 常见问题解答\n\n有关常见问题，请参阅 [FAQ](docs\u002Fen\u002Fnotes\u002Ffaq.md)。\n\n## 贡献说明\n\n我们非常欢迎所有有助于改进 MMDetection3D 的贡献。请参阅 [CONTRIBUTING.md](docs\u002Fen\u002Fnotes\u002Fcontribution_guides.md) 以获取贡献指南。\n\n## 致谢\n\nMMDetection3D 是一个开源项目，由来自不同院校和公司的研究人员与工程师共同贡献而成。我们感谢所有贡献者以及提供宝贵反馈的用户。我们希望这个工具箱和基准测试能够为不断壮大的研究社区服务，提供一个灵活的工具集，用于复现现有方法并开发新的 3D 检测器。\n\n## 引用\n\n如果您在研究中发现本项目有用，请考虑引用：\n\n```latex\n@misc{mmdet3d2020,\n    title={{MMDetection3D: OpenMMLab}下一代通用 3D 物体检测平台},\n    author={MMDetection3D Contributors},\n    howpublished = {\\url{https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d}},\n    year={2020}\n}\n```\n\n## 许可证\n\n本项目采用 [Apache 2.0 许可证](LICENSE) 发布。\n\n## OpenMMLab 中的项目\n\n- [MMEngine](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmengine)：OpenMMLab 用于训练深度学习模型的基础库。\n- [MMCV](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmcv)：OpenMMLab 用于计算机视觉的基础库。\n- [MMEval](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmeval)：一个面向多种机器学习框架的统一评估库。\n- [MIM](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmim)：MIM 用于安装 OpenMMLab 的相关软件包。\n- [MMPreTrain](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmpretrain)：OpenMMLab 的预训练工具箱及基准测试平台。\n- [MMDetection](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection)：OpenMMLab 的目标检测工具箱及基准测试平台。\n- [MMDetection3D](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d)：OpenMMLab 面向通用 3D 物体检测的新一代平台。\n- [MMRotate](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmrotate)：OpenMMLab 的旋转目标检测工具箱及基准测试平台。\n- [MMyOLO](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmyolo)：OpenMMLab 的 YOLO 系列工具箱及基准测试平台。\n- [MMSegmentation](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmsegmentation)：OpenMMLab 的语义分割工具箱及基准测试平台。\n- [MMOCR](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmocr)：OpenMMLab 的文本检测、识别与理解工具箱。\n- [MMPose](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmpose)：OpenMMLab 的人体姿态估计工具箱及基准测试平台。\n- [MMHuman3D](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmhuman3d)：OpenMMLab 的 3D 人体参数化模型工具箱及基准测试平台。\n- [MMSelfSup](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmselfsup)：OpenMMLab 的自监督学习工具箱及基准测试平台。\n- [MMRazor](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmrazor)：OpenMMLab 的模型压缩工具箱及基准测试平台。\n- [MMFewShot](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmfewshot)：OpenMMLab 的小样本学习工具箱及基准测试平台。\n- [MMAction2](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmaction2)：OpenMMLab 的新一代动作理解工具箱及基准测试平台。\n- [MMTracking](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmtracking)：OpenMMLab 的视频感知工具箱及基准测试平台。\n- [MMFlow](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmflow)：OpenMMLab 的光流估计工具箱及基准测试平台。\n- [MMagic](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmagic)：Open**MM**Lab **A**dvanced, **G**enerative and **I**ntelligent **C**reation toolbox。\n- [MMGeneration](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmgeneration)：OpenMMLab 的图像和视频生成模型工具箱。\n- [MMDeploy](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdeploy)：OpenMMLab 的模型部署框架。","# MMDetection3D 快速上手指南\n\nMMDetection3D 是一个基于 PyTorch 的开源 3D 目标检测工具箱，支持多种模态（激光雷达、相机、多模态）及室内外场景的 3D 检测任务。\n\n## 1. 环境准备\n\n在开始之前，请确保您的系统满足以下要求：\n\n*   **操作系统**: Linux (推荐 Ubuntu 18.04+)\n*   **Python**: 3.7+\n*   **PyTorch**: 1.8+ (推荐 2.0+)\n*   **CUDA**: 11.0+ (根据显卡驱动版本选择)\n*   **GCC**: 5.4+\n\n**前置依赖安装建议：**\n建议使用 `conda` 创建虚拟环境，并优先使用国内镜像源加速下载。\n\n```bash\n# 创建虚拟环境\nconda create -n openmmlab python=3.9 -y\nconda activate openmmlab\n\n# 配置国内镜像源 (清华源)\nconda config --add channels https:\u002F\u002Fmirrors.tuna.tsinghua.edu.cn\u002Fanaconda\u002Fpkgs\u002Fmain\u002F\nconda config --add channels https:\u002F\u002Fmirrors.tuna.tsinghua.edu.cn\u002Fanaconda\u002Fpkgs\u002Ffree\u002F\nconda config --set show_channel_urls yes\n```\n\n## 2. 安装步骤\n\n推荐使用 `MIM` (OpenMMLab 包管理工具) 进行安装，它会自动处理复杂的依赖关系。\n\n### 第一步：安装 PyTorch 和 MMCV\n请访问 [PyTorch 官网](https:\u002F\u002Fpytorch.org\u002Fget-started\u002Flocally\u002F) 获取适合您环境的安装命令。以下以 CUDA 11.8 为例：\n\n```bash\n# 使用 pip 安装 (国内用户可添加 -i 参数使用清华源)\npip install torch torchvision torchaudio --index-url https:\u002F\u002Fdownload.pytorch.org\u002Fwhl\u002Fcu118\n\n# 安装 MMCV (MMDetection3D 的核心依赖)\npip install -U openmim\nmim install \"mmcv>=2.0.0\"\n```\n\n### 第二步：安装 MMDetection3D\n您可以选择直接安装发布版或从源码安装（推荐源码安装以便开发）。\n\n**方式 A：使用 pip 安装 (稳定版)**\n```bash\npip install mmdet3d -U\n```\n\n**方式 B：从源码安装 (最新版，推荐)**\n```bash\n# 克隆仓库\ngit clone https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d.git\ncd mmdetection3d\n\n# 安装依赖\npip install -v -e . \n# 注：-v 表示详细输出，-e 表示可编辑模式安装\n```\n\n### 验证安装\n运行以下命令检查是否安装成功：\n```bash\npython -c \"import mmdet3d; print(mmdet3d.__version__)\"\n```\n若无报错并输出版本号，则安装成功。\n\n## 3. 基本使用\n\n以下演示如何使用预训练模型对单个点云文件进行推理（以 PointPillars 模型和 KITTI 数据集格式为例）。\n\n### 准备数据\n确保您有一个 `.bin` 格式的点云文件（例如 `demo.bin`）。如果是首次使用，需下载对应的配置文件和权重文件。\n\n### 执行推理\n使用 `demo\u002Finferencer.py` 或直接调用命令行工具进行推理。\n\n**方法一：使用命令行工具 (最简单)**\n\n```bash\n# 语法：python demo\u002Finferencer.py \u003C点云文件路径> \u003C配置文件路径> \u003C权重文件路径> --out-dir \u003C输出目录>\n\n# 示例：使用 PointPillars 模型推理\npython demo\u002Finferencer.py \\\n    demo\u002Fkitti_sample.bin \\\n    configs\u002Fpointpillars\u002Fhv_pointpillars_secfpn_6x8_160e_kitti-3d-3class.py \\\n    https:\u002F\u002Fdownload.openmmlab.com\u002Fmmdetection3d\u002Fv0.1.0_models\u002Fpointpillars\u002Fhv_pointpillars_secfpn_6x8_160e_kitti-3d-3class\u002Fhv_pointpillars_secfpn_6x8_160e_kitti-3d-3class_20200620_230801-7b5072d0.pth \\\n    --out-dir results\n```\n\n*注：如果无法直接访问 GitHub 上的 config 文件，请先将配置文件下载到本地，然后替换为本地路径。*\n\n**方法二：使用 Python API**\n\n```python\nfrom mmdet3d.apis import MMDet3DInferencer\n\n# 初始化推理器\ninferencer = MMDet3DInferencer(\n    model='configs\u002Fpointpillars\u002Fhv_pointpillars_secfpn_6x8_160e_kitti-3d-3class.py',\n    weights='https:\u002F\u002Fdownload.openmmlab.com\u002Fmmdetection3d\u002Fv0.1.0_models\u002Fpointpillars\u002Fhv_pointpillars_secfpn_6x8_160e_kitti-3d-3class\u002Fhv_pointpillars_secfpn_6x8_160e_kitti-3d-3class_20200620_230801-7b5072d0.pth'\n)\n\n# 执行推理\nresult = inferencer('demo\u002Fkitti_sample.bin', out_dir='results')\n\n# 结果将保存在 results 文件夹中，包含可视化图片和预测数据\n```\n\n### 查看结果\n推理完成后，结果图片将保存在指定的 `--out-dir` 目录中，图片上会标注检测到的 3D 边界框。","某自动驾驶初创团队正在开发城市道路感知系统，急需从激光雷达点云数据中高精度地识别车辆、行人及障碍物。\n\n### 没有 mmdetection3d 时\n- **算法复现困难**：团队需手动从零复现 PointPillars 或 VoxelNet 等经典论文代码，耗时数周且极易出现与原文不一致的 Bug。\n- **数据格式混乱**：不同数据集（如 KITTI、nuScenes）的点云格式差异巨大，需要编写大量重复的脚本进行清洗和对齐。\n- **模型对比低效**：缺乏统一的评估基准，更换骨干网络或调整锚框策略时，难以快速量化性能提升，导致迭代方向模糊。\n- **部署门槛极高**：训练好的模型缺乏标准化的导出接口，移植到车载嵌入式设备时需要耗费大量精力重写推理引擎。\n\n### 使用 mmdetection3d 后\n- **开箱即用模型**：直接调用内置的 20+ 种主流 3D 检测算法和预训练权重，将算法验证周期从数周缩短至几天。\n- **统一数据加载**：利用其标准化的数据接口，一键适配多种公开数据集，彻底消除了繁琐的数据预处理工作。\n- **模块化实验**：通过修改配置文件即可灵活组合不同的骨干网络、颈部结构和损失函数，快速完成多方案性能对比。\n- **顺畅落地部署**：借助完善的模型转换工具链，可轻松将训练好的模型导出为 ONNX 或 TensorRT 格式，加速车端实时推理。\n\nmmdetection3d 通过标准化的全流程框架，让研发团队从重复造轮子中解放出来，专注于核心场景的算法优化与落地。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fopen-mmlab_mmdetection3d_5eb14b53.png","open-mmlab","OpenMMLab","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002Fopen-mmlab_7c171dd7.png","",null,"https:\u002F\u002Fopenmmlab.com","https:\u002F\u002Fgithub.com\u002Fopen-mmlab",[80,84,88,92,96,100],{"name":81,"color":82,"percentage":83},"Python","#3572A5",98.1,{"name":85,"color":86,"percentage":87},"Cuda","#3A4E3A",0.7,{"name":89,"color":90,"percentage":91},"Shell","#89e051",0.6,{"name":93,"color":94,"percentage":95},"C++","#f34b7d",0.4,{"name":97,"color":98,"percentage":99},"MATLAB","#e16737",0.1,{"name":101,"color":102,"percentage":99},"Dockerfile","#384d54",6355,1741,"2026-04-07T04:03:39","Apache-2.0",4,"Linux","需要 NVIDIA GPU (基于 PyTorch\u002FCUDA)，具体型号和显存未说明，需支持 CUDA","未说明",{"notes":112,"python":113,"dependencies":114},"该工具是基于 PyTorch 的 3D 检测工具箱，主分支兼容 PyTorch 1.8+。README 中未直接列出具体的操作系统、Python 版本及硬件数值要求，详细安装步骤需参考官方文档链接。支持多种室内外 3D 检测数据集（如 KITTI, nuScenes, Waymo 等）。","未说明 (依赖 PyTorch 1.8+)",[115,116,117],"torch>=1.8","mmcv","mmdetection",[15,14,119],"其他",[121,122,123,124],"pytorch","3d-object-detection","object-detection","point-cloud","2026-03-27T02:49:30.150509","2026-04-07T18:40:09.432643",[128,133,138,142,147,152],{"id":129,"question_zh":130,"answer_zh":131,"source_url":132},22642,"运行测试或训练时遇到 'KeyError: NuScenesDataset: infos' 错误怎么办？","该错误通常是因为数据集的 info 文件未正确生成或路径配置错误。请确保已按照文档要求预处理数据集（如 NuScenes），并生成了对应的 pkl 信息文件。检查配置文件中的数据根目录和 info 文件路径是否正确指向生成的文件位置。","https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fissues\u002F769",{"id":134,"question_zh":135,"answer_zh":136,"source_url":137},22643,"使用 FCOS3D 或 PGD 模型推理时检测到过多的 3D 边界框（误检严重）如何解决？","检测到过多边界框通常是由于置信度阈值设置过低或后处理参数（如 NMS 阈值）不合适导致的。建议检查配置文件中的 `test_cfg` 部分，适当提高 `score_thr`（置信度阈值）或调整 `nms_thr`。此外，确保使用的权重文件与配置文件匹配，并在验证集上评估以确认超参数是否合理。","https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fissues\u002F1024",{"id":139,"question_zh":140,"answer_zh":141,"source_url":137},22644,"如何批量测试模型并保存所有结果的可视化图像，而不是逐个查看？","在使用 `tools\u002Ftest.py` 进行推理时，添加 `--show-dir` 参数可以将所有批次的预测结果可视化并保存到指定目录，而无需逐个交互查看。命令示例：`python tools\u002Ftest.py \u003Cconfig_file> \u003Ccheckpoint_file> --show --show-dir \u003Coutput_directory>`。生成的文件通常包含原始图像、预测结果和真值标注（如有）。",{"id":143,"question_zh":144,"answer_zh":145,"source_url":146},22645,"在自定义数据集上训练 MVXNet 时遇到 'CUDA error: an illegal memory access was encountered' 错误的原因是什么？","该错误通常是由于修改了点云范围（point_cloud_range）但未同步更新中间特征图的形状（sparse_shape）导致的。计算公式为：特征图尺寸 = (范围最大值 - 范围最小值) \u002F 体素大小 + 1。例如，若点云范围为 [0, -40, -3, 70.4, 40, 1]，体素大小为 [0.05, 0.05, 0.1]，则 Z 轴特征图尺寸为 (1 - (-3)) \u002F 0.1 + 1 = 41。请根据新的点云范围和体素大小重新计算并修改配置文件中的 `voxel_layer` 相关参数。","https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fissues\u002F382",{"id":148,"question_zh":149,"answer_zh":150,"source_url":151},22646,"运行时出现 'ValueError: numpy.ndarray size changed, may indicate binary incompatibility' 错误如何解决？","此错误表明 NumPy 版本与已编译的二进制包（如 pycocotools 或 mmcv）不兼容。通常是因为 NumPy 被升级到了新版本，而其他依赖包是基于旧版本编译的。解决方法是重新安装与当前 NumPy 版本兼容的依赖包，或者降级\u002F升级 NumPy 到与其他包匹配的版本。建议尝试重新安装 `pycocotools` (`pip install --force-reinstall pycocotools`) 或重建整个虚拟环境以确保版本一致性。","https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fissues\u002F650",{"id":153,"question_zh":154,"answer_zh":155,"source_url":137},22647,"推理生成的图片中，'img.png', 'pred.png' 和 'gt.png' 分别代表什么含义？","在使用 `--show` 参数进行推理时，会生成三种图片：\n1. `xxx_img.png`：原始输入图像。\n2. `xxx_pred.png`：模型的预测结果可视化图（通常包含检测框）。\n3. `xxx_gt.png`：真值（Ground Truth）标注可视化图，用于对比预测效果，图中通常显示橙色的标注框。",[157,162,167,172,177,182,187,192,197,202,207,212,217,222,227,232,237,242,247,252],{"id":158,"version":159,"summary_zh":160,"released_at":161},136367,"v1.4.0","## 亮点\n\n- 在 `projects` 中支持 [DSVT](https:\u002F\u002Farxiv.org\u002Fabs\u002F2301.06051) 的训练 (#2738)\n- 在 `projects` 中支持 [Nerf-Det](https:\u002F\u002Farxiv.org\u002Fabs\u002F2307.14620) (#2732)\n- 重构 Waymo 数据集 (#2836)\n\n## 新特性\n\n- 在 `projects` 中支持 [DSVT](https:\u002F\u002Farxiv.org\u002Fabs\u002F2301.06051) 的训练 (#2738)\n- 在 `projects` 中支持 [Nerf-Det](https:\u002F\u002Farxiv.org\u002Fabs\u002F2307.14620) (#2732)\n- 支持 [MV-FCOS3D++](https:\u002F\u002Farxiv.org\u002Fabs\u002F2207.12716) (#2835)\n- 重构 Waymo 数据集 (#2836)\n\n## 改进\n\n- 在 Waymo 数据集中支持 [PGD](https:\u002F\u002Farxiv.org\u002Fabs\u002F2107.14160)（前视图 \u002F 多视角）(#2835)\n- 发布新的 [Waymo-mini](https:\u002F\u002Fdownload.openmmlab.com\u002Fmmdetection3d\u002Fdata\u002Fwaymo_mmdet3d_after_1x4\u002Fwaymo_mini.tar.gz)，用于验证某些方法或快速调试 (#2835)\n\n## Bug 修复\n\n- 修复 MinkUNet 和 SPVCNN 中的一些错误配置 (#2854)\n- 修复 PETR 中参数数量不正确的问题 (#2800)\n- 删除 `mmdet3d\u002Fconfigs` 中的未使用文件 (#2773)\n\n## 贡献者\n\n本次发布共有 5 名开发者参与贡献。\n\n@sunjiahao1999、@WendellZ524、@Yanyirong、@JingweiZhang12、@Tai-Wang\n\n## 新贡献者\n* @WendellZ524 在 https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F2800 中完成了首次贡献\n* @Yanyirong 在 https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F2732 中完成了首次贡献\n\n**完整变更日志**: https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fcompare\u002Fv1.3.0...v1.4.0","2024-01-08T07:52:15",{"id":163,"version":164,"summary_zh":165,"released_at":166},136368,"v1.3.0","## 亮点\n\n- 在 `projects` 中支持 [CENet](https:\u002F\u002Farxiv.org\u002Fabs\u002F2207.12691) (#2619)\n- 使用新的 3D 推理器增强演示示例 (#2763)\n\n## 新特性\n\n- 在 `projects` 中支持 [CENet](https:\u002F\u002Farxiv.org\u002Fabs\u002F2207.12691) (#2619)\n\n## 改进\n\n- 使用新的 3D 推理器增强演示示例 (#2763)\n- 在 nuScenes 数据集教程中添加基于 BEV 的检测流水线 (#2672)\n- 在 `mmdet3d\u002Fconfigs` 中新增 Cylinder3D 配置类型 (#2681)\n- 更新 [新配置类型](https:\u002F\u002Fmmengine.readthedocs.io\u002Fen\u002Flatest\u002Fadvanced_tutorials\u002Fconfig.html#a-pure-python-style-configuration-file-beta) (#2655)\n- 更新 README.md 中的二维码 (#2703)\n\n## Bug 修复\n\n- 修复 nuScenes 数据集的下载脚本 (#2660)\n- 修复 circleCI 和 GitHub 工作流配置 (#2652)\n- 修复 requirements 中 Open3D 的版本问题 (#2633)\n- 修复 `mmdet3d\u002Fconfigs` 中的未使用文件 (#2773)\n- 修复 FreeAnchor3DHead 中对设备的支持问题 (#2769)\n- 修复 ReadTheDocs 构建及链接问题 (#2739, #2650)\n- 修复 LaserMix 中的俯仰角 bug (#2710)\n\n## 贡献者\n\n本次发布共有 7 名开发者做出贡献。\n\n@sunjiahao1999、@Xiangxu-0103、@ZhaoCake、@LRJKD、@crazysteeaam、@wep21、@zhiqwang\n\n## 新贡献者\n* @wep21 在 https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F2660 中完成了首次贡献。\n* @zhiqwang 在 https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F2650 中完成了首次贡献。\n* @1uciusy 在 https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F2672 中完成了首次贡献。\n* @crazysteeaam 在 https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F2703 中完成了首次贡献。\n* @ZhaoCake 在 https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F2681 中完成了首次贡献。\n* @LRJKD 在 https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F2769 中完成了首次贡献。\n\n**完整变更日志**: https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fcompare\u002Fv1.2.0...v1.3.0","2023-10-19T07:42:13",{"id":168,"version":169,"summary_zh":170,"released_at":171},136369,"v1.2.0","## 亮点\n\n- 在 `mmdet3d\u002Fconfig` 中支持[新的配置类型](https:\u002F\u002Fmmengine.readthedocs.io\u002Fen\u002Flatest\u002Fadvanced_tutorials\u002Fconfig.html#a-pure-python-style-configuration-file-beta) (#2608)\n- 在 `projects` 中支持 [DSVT](\u003C(https:\u002F\u002Farxiv.org\u002Fabs\u002F2301.06051)>) 的推理 (#2606)\n- 支持使用 `mim` 从 [OpenDataLab](https:\u002F\u002Fopendatalab.com\u002F) 下载数据集 (#2593)\n\n## 新特性\n\n- 在 `mmdet3d\u002Fconfig` 中支持[新的配置类型](https:\u002F\u002Fmmengine.readthedocs.io\u002Fen\u002Flatest\u002Fadvanced_tutorials\u002Fconfig.html#a-pure-python-style-configuration-file-beta) (#2608)\n- 在 `projects` 中支持 [DSVT](\u003C(https:\u002F\u002Farxiv.org\u002Fabs\u002F2301.06051)>) 的推理 (#2606)\n- 支持使用 `mim` 从 [OpenDataLab](https:\u002F\u002Fopendatalab.com\u002F) 下载数据集 (#2593)\n\n## 改进\n\n- 增强了交互式可视化功能 (#2611)\n- 更新了 README.md 和模型库 (#2599, #2600)\n- 加快了 S3DIS 数据准备的速度 (#2585)\n\n## Bug 修复\n\n- 移除基准训练中的 PointRCNN (#2610)\n- 修复了室内检测可视化错误 (#2625)\n- 修复了 MinkUNet 的下载链接 (#2590)\n- 修正了 `readthedocs` 中的公式 (#2580)\n\n## 贡献者\n\n本次发布共有 5 名开发者参与贡献。\n\n@sunjiahao1999、@Xiangxu-0103、@JingweiZhang12、@col14m、@zhulf0804\n\n## 新贡献者\n* @col14m 在 https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F2585 中完成了首次贡献\n* @zhulf0804 在 https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F2600 中完成了首次贡献\n\n**完整变更日志**: https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fcompare\u002Fv1.1.1...v1.2.0","2023-07-04T15:51:17",{"id":173,"version":174,"summary_zh":175,"released_at":176},136370,"v1.1.1","## 亮点\n\n- 在 `projects` 中支持 [TPVFormer](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2302.07817.pdf) (#2399, #2517, #2535)\n- 在 `projects` 中支持 BEVFusion 的训练 (#2546)\n- 支持基于激光雷达的 3D 语义分割基准测试 (#2530, #2559)\n\n## 新特性\n\n- 在 `projects` 中支持 [TPVFormer](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2302.07817.pdf) (#2399, #2517, #2535)\n- 在 `projects` 中支持 BEVFusion 的训练 (#2558)\n- 支持基于激光雷达的 3D 语义分割基准测试 (#2530, #2559)\n- 支持 Segmentor 的测试时增强 (#2382)\n- 支持 `Minkowski ConvModule` 和 `Residual` 块 (#2528)\n- 支持在多模态方法中可视化多视角图像 (#2453)\n\n## 改进\n\n- 上传 PETR 的检查点和训练日志 (#2555)\n- 在分割评估中将 `np.float` 默认替换为 `float` (#2527)\n- 添加 SemanticKITTI 数据集转换文档 (#2515)\n- 在可视化中支持不同类别使用不同颜色 (#2500)\n- 为 `BaseInstance3DBoxes` 和 `BasePoint` 支持类似张量的操作 (#2501)\n- 在 NuScenes 标注文件中添加激光雷达分割信息\n- 提供离线生成的数据集标注文件 (#2457)\n- 重构文档结构 (#2429)\n- 完成类型提示和文档字符串 (#2396, #2457, #2468, #2464, #2485)\n\n## Bug 修复\n\n- 修复在自动混合精度 (AMP) 模式下训练 SECOND 时损失异常的 bug (#2452)\n- 在 mmdet3d\u002Fdataset\u002Fconvert_utils.py 的 `post_process_coords` 函数中添加警告 (#2557)\n- 修复无效的配置文件 (#2477, #2536)\n- 修复单元测试中的错误 (#2466)\n- 更新 test.py 中的 `local-rank` 参数以兼容 PyTorch 2.0 (#2469)\n- 修复 Docker 文件 (#2451)\n- 修复演示和可视化相关问题 (#2453)\n- 修复 SUN RGB-D 数据转换器 (#2440)\n- 修复 ReadTheDocs 构建问题 (#2459, #2419, #2505, #2396)\n- 修复 CI 问题 (#2445, #2424)\n- 放宽 `numba` 的版本限制 (#2416)\n\n## 贡献者\n\n本次发布共有 12 名开发者参与贡献。\n\n@sunjiahao1999, @Xiangxu-0103, @JingweiZhang12, @chriscarving, @jaan1729, @pd-michaelstanley, @filaPro, @kabouzeid, @A-new-b, @lbin, @Lum1104, @pd-michaelstanley\n\n\n## 新贡献者\n* @A-new-b 在 https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F2485 中完成了首次贡献\n* @lbin 在 https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F2557 中完成了首次贡献\n* @Lum1104 在 https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F2464 中完成了首次贡献\n* @jaan1729 在 https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F2451 中完成了首次贡献\n* @pd-michaelstanley 在 https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F2442 中完成了首次贡献\n* @chriscarving 在 https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F2396 中完成了首次贡献\n\n**完整变更日志**: https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fcompare\u002Fv1.1.0...v1.1.1","2023-05-31T07:57:04",{"id":178,"version":179,"summary_zh":180,"released_at":181},136371,"v1.1.0","我们很高兴地宣布，作为 OpenMMLab 2.0 项目的一部分，**MMDetection3D 1.1.0** 已正式发布！与 1.0.0 版本相比，1.1.0 引入了核心包的更新框架结构，并新增了一个名为 `Projects` 的模块。具体而言，我们对核心包的代码进行了大幅重构，使其更加清晰易懂、模块化程度更高。新的 `Projects` 模块是 MMDetection3D 的重要补充，它支持灵活的代码贡献方式，无需严格的代码规范约束，从而能够更快地集成最前沿的模型和功能。为了帮助用户尽可能平滑地从 1.0.0 迁移到 1.1.0，我们准备了一份[迁移指南](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fblob\u002Fmain\u002Fdocs\u002Fen\u002Fmigration.md)。如有关于迁移的任何疑问，欢迎在 [issue](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fissues) 中提出。\n\n\n## 亮点\n\n- 支持多种流行的 LiDAR 分割方法：\n    + [Cylinder3D](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2011.10033.pdf) (#2291, #2344, #2350)\n    + [MinkUnet](https:\u002F\u002Farxiv.org\u002Fabs\u002F1904.08755) (#2294, #2358) \n    + [SPVCNN](https:\u002F\u002Farxiv.org\u002Fabs\u002F2007.16100) (#2320，#2372)\n    + [PolarMix](https:\u002F\u002Farxiv.org\u002Fabs\u002F2208.00223) 数据增强 (#2265)\n    + [LaserMix](https:\u002F\u002Farxiv.org\u002Fabs\u002F2207.00026) 数据增强 (#2302)\n- 在 `projects` 模块中支持 [TR3D](https:\u002F\u002Farxiv.org\u002Fabs\u002F2302.02858) 检测器 (#2274)\n- 在 `projects` 模块中支持 [DETR3D](https:\u002F\u002Farxiv.org\u002Fabs\u002F2110.06922) (#2173)\n- 在 `projects` 模块中支持 [BEVFusion](https:\u002F\u002Farxiv.org\u002Fabs\u002F2205.13542) 的推理 (#2175)\n\n## 新特性\n\n- 支持 [Cylinder3D](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2011.10033.pdf) (#2291, #2344, #2350)\n- 支持 [MinkUnet](https:\u002F\u002Farxiv.org\u002Fabs\u002F1904.08755) (#2294, #2358)\n- 支持 [SPVCNN](https:\u002F\u002Farxiv.org\u002Fabs\u002F2007.16100) (#2320，#2372)\n- 在 `projects` 模块中支持 [TR3D](https:\u002F\u002Farxiv.org\u002Fabs\u002F2302.02858) 检测器 (#2274)\n- 在 `projects` 模块中支持 [BEVFusion](https:\u002F\u002Farxiv.org\u002Fabs\u002F2205.13542) 的推理 (#2175)\n- 在 `projects` 模块中支持 [DETR3D](https:\u002F\u002Farxiv.org\u002Fabs\u002F2110.06922) (#2173)\n- 支持 [PolarMix](https:\u002F\u002Farxiv.org\u002Fabs\u002F2208.00223) 和 [LaserMix](https:\u002F\u002Farxiv.org\u002Fabs\u002F2207.00026) 数据增强 (#2265,#2302)\n- 支持加载全景分割标注 (#2223)\n- 支持全景分割评估指标 (#2230)\n- 新增基于 LiDAR、单目及多模态 3D 检测的推理器 (#2208, #2190, #2342)\n- 新增基于 LiDAR 的分割推理器 (#2304)\n\n## 改进\n\n- 支持 CBGSDataset 的 `lazy_init` 功能 (#2271)\n- 支持为 Waymo 数据集生成测试集标注文件 (#2180)\n- 加强对 SemanticKitti 的支持 (#2253, #2323)\n- 文件 I\u002FO 的迁移与重构 (#2319)\n- 支持 Lyft、NuScenes 和 Waymo 数据集的 `format_only` 选项 (#2333, #2151)\n- 使用 `torch.permute` 替代 `np.transpose` 以提升速度 (#2277)\n- 允许为 PyTorch 2.0 设置 local-rank (#2387)\n\n## Bug 修复\n\n- 修复了 le 反转的问题","2023-04-19T05:57:10",{"id":183,"version":184,"summary_zh":185,"released_at":186},136372,"v1.1.0rc3","## 亮点\n\n- 在 `projects` 中支持 [CenterFormer](https:\u002F\u002Farxiv.org\u002Fabs\u002F2209.05588) (#2175)\n- 在 `projects` 中支持 [PETR](https:\u002F\u002Farxiv.org\u002Fabs\u002F2203.05625) (#2173)\n\n## 新特性\n\n- 在 `projects` 中支持 [CenterFormer](https:\u002F\u002Farxiv.org\u002Fabs\u002F2209.05588) (#2175)\n- 在 `projects` 中支持 [PETR](https:\u002F\u002Farxiv.org\u002Fabs\u002F2203.05625) (#2173)\n- 将 SUN RGB-D 数据集上的 ImVoxelNet 重构至 mmdet3d v1.1 (#2141)\n\n## 改进\n\n- 移除旧版的 builder.py (#2061)\n- 更新 `customize_dataset` 文档 (#2153)\n- 更新基于 LiDAR 的目标检测教程 (#2120)\n\n## Bug 修复\n\n- 修复 FCOS3D 和 PGD 的配置文件 (#2191)\n- 修复 update_infos_to_v2.py 中 NumPy 的 `ValueError` 异常 (#2162)\n- 修复 Det3DVisualizationHook 中缺失的参数 (#2118)\n- 修复旋转框 IoU 计算中的内存溢出问题 (#2134)\n- 修复 update_infos_to_v2.py 中 NUS 和 Lyft 数据集的 lidar2cam 转换错误 (#2110)\n- 修复 Waymo 评估指标中数据类型错误的问题 (#2109)\n- 更新单目 3D 目标检测任务中 `cam_instances` 的 `bbox_3d` 信息 (#2046)\n- 修复 Waymo 数据集标签保存问题 (#2096)\n\n## 贡献者\n\n本次发布共有 10 名开发者参与贡献。\n\n@SekiroRong、@ZLTJohn、@vansin、@shanmo、@VVsssssk、@ZCMax、@Xiangxu-0103、@JingweiZhang12、@Tai-Wang、@lianqing11\n\n## 新贡献者\n* @shanmo 在 https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F2118 中完成了首次贡献\n* @ZLTJohn 在 https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F2162 中完成了首次贡献\n* @SekiroRong 在 https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F2175 中完成了首次贡献\n\n**完整变更日志**: https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fcompare\u002Fv1.1.0rc2...v1.1.0rc3","2023-01-10T03:23:09",{"id":188,"version":189,"summary_zh":190,"released_at":191},136373,"v1.0.0rc6","### 新特性\n\n- 添加 `Projects\u002F` 文件夹及首个示例项目 (#2082)\n\n### 改进\n\n- 更新 Waymo 转换器以节省存储空间 (#1759)\n- 更新 CenterPoint 的模型链接及性能指标 (#1916)\n\n### Bug 修复\n\n- 修复 PointRCNN 中的 GPU 显存占用问题 (#1928)\n- 修复 `IoUNegPiecewiseSampler` 中的采样 bug (#2018)\n\n### 贡献者\n\n共有 6 名开发者参与了本次发布。\n\n@oyel, @zzj403, @VVsssssk, @Tai-Wang, @tpoisonooo, @JingweiZhang12, @ZCMax, @ZwwWayne\n\n\n**完整变更日志**: https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fcompare\u002Fv1.0.0rc5...v1.0.0rc6","2022-12-16T12:22:07",{"id":193,"version":194,"summary_zh":195,"released_at":196},136374,"v1.1.0rc2","## 亮点\n\n- 支持 [PV-RCNN](https:\u002F\u002Farxiv.org\u002Fabs\u002F1912.13192)\n- 加速 Waymo 数据集上的评估\n\n## 新特性\n\n- 支持 [PV-RCNN](https:\u002F\u002Farxiv.org\u002Fabs\u002F1912.13192) (#1597, #2045)\n- 加速 Waymo 数据集上的评估 (#2008)\n- 将 FCAF3D 重构到 mmdet3d v1.1 的框架中 (#1945)\n- 将 S3DIS 数据集重构到 mmdet3d v1.1 的框架中 (#1984)\n- 添加 `Projects\u002F` 文件夹及首个示例项目 (#2042)\n\n## 改进\n\n- 将 `CLASSES` 和 `PALETTE` 分别重命名为 `classes` 和 `palette` (#1932)\n- 更新 pkl 文件中的 `metainfo`，并在 `metainfo` 中添加 `categories` (#1934)\n- 在管道前后展示实例统计信息 (#1863)\n- 为不同测试区域添加 DGCNN 的配置文件 (#1967)\n- 将测试工具从 `tests\u002Futils\u002F` 移至 `mmdet3d\u002Ftesting\u002F` (#2012)\n- 为 `models\u002Flayers\u002F` 中的代码添加类型提示 (#2014)\n- 完善文档 (#1891, #1994)\n- 优化体素化以提升速度 (#2062)\n\n## Bug 修复\n\n- 修复点云循环可视化错误 (#1914)\n- 修复 Waymo 图像转换问题，避免信息丢失 (#1979)\n- 修复 KITTI 测试集上的评估问题 (#2005)\n- 修复 `IoUNegPiecewiseSampler` 中的采样 bug (#2017)\n- 修复 CenterPoint 中的点云范围问题 (#1998)\n- 修复部分加载错误，并支持 Waymo 数据集中基于 FOV 图像的模式 (#1942)\n- 修复数据集转换工具 (#1923, #2040, #1971)\n- 更新所有配置文件中的元文件 (#2006)\n\n## 贡献者\n\n本次发布共有 12 名开发者参与贡献。\n\n@vavanade、@oyel、@thinkthinking、@PeterH0323、@274869388、@cxiang26、@lianqing11、@VVsssssk、@ZCMax、@Xiangxu-0103、@JingweiZhang12、@Tai-Wang\n\n## 新贡献者\n\n- @PeterH0323 在 #2065 中完成了首次贡献\n- @cxiang26 在 #1965 中完成了首次贡献\n- @vavanade 在 #2031 中完成了首次贡献\n- @oyel 在 #2017 中完成了首次贡献\n- @thinkthinking 在 #2026 中完成了首次贡献\n- @274869388 在 #1973 中完成了首次贡献\n\n**完整变更日志**: https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fcompare\u002Fv1.1.0rc1...v1.1.0rc2","2022-12-03T12:59:16",{"id":198,"version":199,"summary_zh":200,"released_at":201},136375,"v1.1.0rc1","### 亮点\n\n- 支持在 Waymo 数据集上运行纯相机 3D 检测基线 [MV-FCOS3D++](https:\u002F\u002Farxiv.org\u002Fabs\u002F2207.12716)\n\n### 新特性\n\n- 支持在 Waymo 数据集上运行纯相机 3D 检测基线 [MV-FCOS3D++](https:\u002F\u002Farxiv.org\u002Fabs\u002F2207.12716)，并新增评估指标和数据变换功能 (#1716)\n- 在 mmdet3d v1.1 的框架下重构 PointRCNN (#1819)\n\n### 改进\n\n- 在配置文件中添加 `auto_scale_lr`，以支持自动缩放学习率的训练 (#1807)\n- 修复 CI 相关问题 (#1813、#1865、#1877)\n- 更新 `browse_dataset.py` 脚本 (#1817)\n- 更新 SUN RGB-D 和 Lyft 数据集的文档 (#1833)\n- 将检测器中的 `convert_to_datasample` 重命名为 `add_pred_to_datasample` (#1843)\n- 更新自定义数据集的文档 (#1845)\n- 更新 `Det3DLocalVisualization` 及可视化相关文档 (#1857)\n- 添加生成 Waymo 数据集 `cam_sync_labels` 的代码 (#1870)\n- 更新数据集变换的类型提示 (#1875)\n\n### Bug 修复\n\n- 修复 [setup_env.py](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fblob\u002Fdev-1.x\u002Fmmdet3d\u002Futils\u002Fsetup_env.py) 中模型注册缺失的问题 (#1808)\n- 修复使用地面平面数据时数据采样器的 bug (#1812)\n- 在可视化过程中增加输出目录是否存在检查 (#1828)\n- 修复 nuScenes 数据集在单目 3D 检测中的 bug (#1837)\n- 修复可视化钩子，使其支持不同模态数据的可视化 (#1839)\n- 修复单目 3D 检测演示程序 (#1864)\n- 修复 nuScenes 数据集中缺少 `num_pts_feats` 键的问题，并完善文档字符串 (#1882)\n\n### 贡献者\n\n本次发布共有 10 位开发者参与贡献。\n\n@ZwwWayne、@Tai-Wang、@lianqing11、@VVsssssk、@ZCMax、@Xiangxu-0103、@JingweiZhang12、@tpoisonooo、@ice-tong、@jshilong\n\n## 新贡献者\n* @ice-tong 在 https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F1838 中完成了首次贡献\n\n**完整变更日志**: https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fcompare\u002Fv1.1.0rc0...v1.1.0rc1","2022-10-17T11:36:09",{"id":203,"version":204,"summary_zh":205,"released_at":206},136376,"v1.0.0rc5","### 新特性\n\n- 支持 SUN RGB-D 数据集上的 ImVoxelNet (#1738)\n\n### 改进\n\n- 修复元文件 README 中的跨代码库引用问题 (#1644)\n- 更新中文入门文档 (#1715)\n- 修复文档链接并添加文档链接检查工具 (#1811)\n\n### Bug 修复\n\n- 修复可能由空预测标签触发的可视化错误 (#1725)\n- 修复因参数传递错误导致的点云分割可视化错误 (#1858)\n- 修复 PointRCNN 训练过程中出现的 NaN 损失问题 (#1874)\n\n### 贡献者\n\n本次发布共有 11 名开发者参与贡献。\n\n@ZwwWayne、@Tai-Wang、@filaPro、@VVsssssk、@ZCMax、@Xiangxu-0103、@holtvogt、@tpoisonooo、@lianqing01、@TommyZihao、@aditya9710\n\n## 新贡献者\n\n* @tpoisonooo 在 https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F1614 中完成了首次贡献\n* @holtvogt 在 https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F1725 中完成了首次贡献\n* @TommyZihao 在 https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F1778 中完成了首次贡献\n* @aditya9710 在 https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F1889 中完成了首次贡献\n\n**完整变更日志**: https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fcompare\u002Fv1.0.0rc4...v1.0.0rc5","2022-10-17T11:33:18",{"id":208,"version":209,"summary_zh":210,"released_at":211},136377,"v1.1.0rc0","# Changelog of v1.1\r\n\r\n### v1.1.0rc0 (1\u002F9\u002F2022)\r\n\r\nWe are excited to announce the release of MMDetection3D 1.1.0rc0.\r\nMMDet3D 1.1.0rc0 is the first version of MMDetection3D 1.1, a part of the OpenMMLab 2.0 projects.\r\nBuilt upon the new [training engine](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmengine) and [MMDet 3.x](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection\u002Ftree\u002F3.x),\r\nMMDet3D 1.1 unifies the interfaces of dataset, models, evaluation, and visualization with faster training and testing speed.\r\nIt also provides a standard data protocol for different datasets, modalities, and tasks for 3D perception.\r\nWe will support more strong baselines in the future release, with our latest exploration on camera-only 3D detection from videos.\r\n\r\n### Highlights\r\n\r\n1. **New engines**. MMDet3D 1.1 is based on [MMEngine](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmengine) and [MMDet 3.x](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection\u002Ftree\u002F3.x), which provides a universal and powerful runner that allows more flexible customizations and significantly simplifies the entry points of high-level interfaces.\r\n\r\n2. **Unified interfaces**. As a part of the OpenMMLab 2.0 projects, MMDet3D 1.1 unifies and refactors the interfaces and internal logics of train, testing, datasets, models, evaluation, and visualization. All the OpenMMLab 2.0 projects share the same design in those interfaces and logics to allow the emergence of multi-task\u002Fmodality algorithms.\r\n\r\n3. **Standard data protocol for all the datasets, modalities, and tasks for 3D perception**. Based on the unified base datasets inherited from MMEngine, we also design a standard data protocol that defines and unifies the common keys across different datasets, tasks, and modalities. It significantly simplifies the usage of multiple datasets and data modalities for multi-task frameworks and eases dataset customization. Please refer to the [documentation of customized datasets](..\u002Fadvanced_guides\u002Fcustomize_dataset.md) for details.\r\n\r\n4. **Strong baselines**. We will release strong baselines of many popular models to enable fair comparisons among state-of-the-art models.\r\n\r\n5. **More documentation and tutorials**. We add a bunch of documentation and tutorials to help users get started more smoothly. Read it [here](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002F1.1\u002F).\r\n\r\n### Breaking Changes\r\n\r\nMMDet3D 1.1 has undergone significant changes to have better design, higher efficiency, more flexibility, and more unified interfaces.\r\nBesides the changes of API, we briefly list the major breaking changes in this section.\r\nWe will update the [migration guide](..\u002Fmigration.md) to provide complete details and migration instructions.\r\nUsers can also refer to the [compatibility documentation](.\u002Fcompatibility.md) and [API doc](https:\u002F\u002Fmmdetection3d.readthedocs.io\u002Fen\u002F1.1\u002F) for more details.\r\n\r\n#### Dependencies\r\n\r\n- MMDet3D 1.1 runs on PyTorch>=1.6. We have deprecated the support of PyTorch 1.5 to embrace the mixed precision training and other new features since PyTorch 1.6. Some models can still run on PyTorch 1.5, but the full functionality of MMDet3D 1.1 is not guaranteed.\r\n- MMDet3D 1.1 relies on MMEngine to run. MMEngine is a new foundational library for training deep learning models of OpenMMLab and are widely depended by OpenMMLab 2.0 projects. The dependencies of file IO and training are migrated from MMCV 1.x to MMEngine.\r\n- MMDet3D 1.1 relies on MMCV>=2.0.0rc0. Although MMCV no longer maintains the training functionalities since 2.0.0rc0, MMDet3D 1.1 relies on the data transforms, CUDA operators, and image processing interfaces in MMCV. Note that the package `mmcv` is the version that provides pre-built CUDA operators and `mmcv-lite` does not since MMCV 2.0.0rc0, while `mmcv-full` has been deprecated since 2.0.0rc0.\r\n- MMDet3D 1.1 is based on MMDet 3.x, which is also a part of OpenMMLab 2.0 projects.\r\n\r\n#### Training and testing\r\n\r\n- MMDet3D 1.1 uses Runner in [MMEngine](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmengine) rather than that in MMCV. The new Runner implements and unifies the building logic of dataset, model, evaluation, and visualizer. Therefore, MMDet3D 1.1 no longer relies on the building logics of those modules in `mmdet3d.train.apis` and `tools\u002Ftrain.py`. Those code have been migrated into [MMEngine](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmengine\u002Fblob\u002Fmain\u002Fmmengine\u002Frunner\u002Frunner.py). Please refer to the [migration guide of Runner in MMEngine](https:\u002F\u002Fmmengine.readthedocs.io\u002Fen\u002Flatest\u002Fmigration\u002Frunner.html) for more details.\r\n- The Runner in MMEngine also supports testing and validation. The testing scripts are also simplified, which has similar logic as that in training scripts to build the runner.\r\n- The execution points of hooks in the new Runner have been enriched to allow more flexible customization. Please refer to the [migration guide of Hook in MMEngine](https:\u002F\u002Fmmengine.readthedocs.io\u002Fen\u002Flatest\u002Fmigration\u002Fhook.html) for more details.\r\n- Learning rate and momentum scheduling has been migrated from Hook to [Parameter Sch","2022-09-01T13:27:48",{"id":213,"version":214,"summary_zh":215,"released_at":216},136378,"v1.0.0rc4","### Highlights\r\n\r\n- Support [FCAF3D](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2112.00322.pdf)\r\n\r\n### New Features\r\n\r\n- Support [FCAF3D](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2112.00322.pdf) (#1547)\r\n- Add the transformation to support multi-camera 3D object detection (#1580)\r\n- Support lift-splat-shoot view transformer (#1598)\r\n\r\n### Improvements\r\n\r\n- Remove the limitation of the maximum number of points during SUN RGB-D preprocessing (#1555)\r\n- Support circle CI (#1647)\r\n- Add mim to extras_require in setup.py (#1560, #1574)\r\n- Update dockerfile package version (#1697)\r\n\r\n### Bug Fixes\r\n\r\n- Flip yaw angle for DepthInstance3DBoxes.overlaps (#1548, #1556)\r\n- Fix DGCNN configs (#1587)\r\n- Fix bbox head not registered bug (#1625)\r\n- Fix missing objects in S3DIS preprocessing (#1665)\r\n- Fix spconv2.0 model loading bug (#1699)\r\n\r\n### Contributors\r\n\r\nA total of 9 developers contributed to this release.\r\n\r\n@Tai-Wang, @ZwwWayne, @filaPro, @lianqing11, @ZCMax, @HuangJunJie2017, @Xiangxu-0103, @ChonghaoSima, @VVsssssk\r\n\r\n## New Contributors\r\n* @HuangJunJie2017 made their first contribution in https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F1580\r\n* @ChonghaoSima made their first contribution in https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F1614\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fcompare\u002Fv1.0.0rc3...v1.0.0rc4","2022-08-08T12:00:00",{"id":218,"version":219,"summary_zh":220,"released_at":221},136379,"v1.0.0rc3","### Highlights\r\n\r\n- Support [SA-SSD](https:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent_CVPR_2020\u002Fpapers\u002FHe_Structure_Aware_Single-Stage_3D_Object_Detection_From_Point_Cloud_CVPR_2020_paper.pdf)\r\n\r\n### New Features\r\n\r\n- Support [SA-SSD](https:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent_CVPR_2020\u002Fpapers\u002FHe_Structure_Aware_Single-Stage_3D_Object_Detection_From_Point_Cloud_CVPR_2020_paper.pdf) (#1337)\r\n\r\n### Improvements\r\n\r\n- Add Chinese documentation for vision-only 3D detection (#1438)\r\n- Update CenterPoint pretrained models that are compatible with refactored coordinate systems (#1450)\r\n- Configure myst-parser to parse anchor tag in the documentation (#1488)\r\n- Replace markdownlint with mdformat for avoiding installing ruby (#1489)\r\n- Add missing `gt_names` when getting annotation info in Custom3DDataset (#1519)\r\n- Support S3DIS full ceph training (#1542)\r\n- Rewrite the installation and FAQ documentation (#1545)\r\n\r\n### Bug Fixes\r\n\r\n- Fix the incorrect registry name when building RoI extractors (#1460)\r\n- Fix the potential problems caused by the registry scope update when composing pipelines (#1466) and using CocoDataset (#1536)\r\n- Fix the missing selection with `order` in the [box3d_nms](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fblob\u002Fmaster\u002Fmmdet3d\u002Fcore\u002Fpost_processing\u002Fbox3d_nms.py) introduced by [#1403](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F1403) (#1479)\r\n- Update the [PointPillars config](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fblob\u002Fmaster\u002Fconfigs\u002Fpointpillars\u002Fhv_pointpillars_secfpn_6x8_160e_kitti-3d-car.py) to make it consistent with the log (#1486)\r\n- Fix heading anchor in documentation (#1490)\r\n- Fix the compatibility of mmcv in the dockerfile (#1508)\r\n- Make overwrite_spconv packaged when building whl (#1516)\r\n- Fix the requirement of mmcv and mmdet (#1537)\r\n- Update configs of PartA2 and support its compatibility with spconv 2.0 (#1538)\r\n\r\n### Contributors\r\n\r\nA total of 13 developers contributed to this release.\r\n\r\n@Xiangxu-0103, @ZCMax, @jshilong, @filaPro, @atinfinity, @Tai-Wang, @wenbo-yu, @yi-chen-isuzu, @ZwwWayne, @wchen61, @VVsssssk, @AlexPasqua, @lianqing11\r\n\r\n## New Contributors\r\n* @atinfinity made their first contribution in https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F1508\r\n* @wenbo-yu made their first contribution in https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F1337\r\n* @wchen61 made their first contribution in https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F1516\r\n* @AlexPasqua made their first contribution in https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F1519\r\n* @lianqing11 made their first contribution in https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F1545\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fcompare\u002Fv1.0.0rc2...v1.0.0rc3","2022-06-14T15:35:03",{"id":223,"version":224,"summary_zh":225,"released_at":226},136380,"v1.0.0rc2","## Highlights\r\n\r\n- Support spconv 2.0\r\n- Support MinkowskiEngine with MinkResNet\r\n- Support training models on custom datasets with only point clouds\r\n- Update Registry to distinguish the scope of built functions\r\n- Replace mmcv.iou3d with a set of bird-eye-view (BEV) operators to unify the operations of rotated boxes\r\n\r\n## New Features\r\n\r\n- Add loader arguments in the configuration files (#1388)\r\n- Support [spconv 2.0](https:\u002F\u002Fgithub.com\u002Ftraveller59\u002Fspconv) when the package is installed. Users can still use spconv 1.x in MMCV with CUDA 9.0 (only cost more memory) without losing the compatibility of model weights between two versions (#1421)\r\n- Support MinkowskiEngine with MinkResNet (#1422)\r\n\r\n## Improvements\r\n\r\n- Add the documentation for model deployment (#1373, #1436)\r\n- Add Chinese documentation of\r\n  - Speed benchmark (#1379)\r\n  - LiDAR-based 3D detection (#1368)\r\n  - LiDAR 3D segmentation (#1420)\r\n  - Coordinate system refactoring (#1384)\r\n- Support training models on custom datasets with only point clouds (#1393)\r\n- Replace mmcv.iou3d with a set of bird-eye-view (BEV) operators to unify the operations of rotated boxes (#1403, #1418)\r\n- Update Registry to distinguish the scope of building functions (#1412, #1443)\r\n- Replace recommonmark with myst_parser for documentation rendering (#1414)\r\n\r\n## Bug Fixes\r\n\r\n- Fix the show pipeline in the [browse_dataset.py](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fblob\u002Fmaster\u002Ftools\u002Fmisc\u002Fbrowse_dataset.py) (#1376)\r\n- Fix missing __init__ files after coordinate system refactoring (#1383)\r\n- Fix the incorrect yaw in the visualization caused by coordinate system refactoring (#1407)\r\n- Fix `NaiveSyncBatchNorm1d` and `NaiveSyncBatchNorm2d` to support non-distributed cases and more general inputs (#1435)\r\n\r\n## Contributors\r\n\r\nA total of 11 developers contributed to this release.\r\n\r\n@ZCMax, @ZwwWayne, @Tai-Wang, @VVsssssk, @HanaRo, @JoeyforJoy, @ansonlcy, @filaPro, @jshilong, @Xiangxu-0103, @deleomike\r\n\r\n## New Contributors\r\n* @HanaRo made their first contribution in https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F1379\r\n* @JoeyforJoy made their first contribution in https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F1368\r\n* @ansonlcy made their first contribution in https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F1391\r\n* @deleomike made their first contribution in https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F1383\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fcompare\u002Fv1.0.0rc1...v1.0.0rc2","2022-05-02T05:41:33",{"id":228,"version":229,"summary_zh":230,"released_at":231},136381,"v1.0.0rc1","## Compatibility\r\n\r\n- We migrate all the mmdet3d ops to mmcv and do not need to compile them when installing mmdet3d.\r\n- To fix the imprecise timestamp and optimize its saving method, we reformat the point cloud data during Waymo data conversion. The data conversion time is also optimized significantly by supporting parallel processing. Please re-generate KITTI format Waymo data if necessary. See more details in the [compatibility documentation](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fblob\u002Fmaster\u002Fdocs\u002Fen\u002Fcompatibility.md).\r\n- We update some of the model checkpoints after the refactor of coordinate systems. Please stay tuned for the release of the remaining model checkpoints.\r\n\r\n|        | Fully Updated   | Partially Updated  |  In Progress  | No Influcence |\r\n|--------------------|:-------------:|:--------:| :-----------: | :-----------: |\r\n| SECOND                  |          | ✓        |               |               |\r\n| PointPillars            |          | ✓        |               |               |\r\n| FreeAnchor              | ✓        |          |               |               |\r\n| VoteNet                 | ✓        |          |               |               |\r\n| H3DNet                  | ✓        |          |               |               |\r\n| 3DSSD                   |          | ✓        |               |               |\r\n| Part-A2                 | ✓        |          |               |               |\r\n| MVXNet                  | ✓        |          |               |               |\r\n| CenterPoint             |          |          |✓              |               |\r\n| SSN                     | ✓        |          |               |               |\r\n| ImVoteNet               | ✓        |          |               |               |\r\n| FCOS3D                  |          |          |               |✓              |\r\n| PointNet++              |          |          |               |✓              |\r\n| Group-Free-3D           |          |          |               |✓              |\r\n| ImVoxelNet              | ✓        |          |               |               |\r\n| PAConv                  |          |          |               |✓              |\r\n| DGCNN                   |          |          |               |✓              |\r\n| SMOKE                   |          |          |               |✓              |\r\n| PGD                     |          |          |               |✓              |\r\n| MonoFlex                |          |          |               |✓              |\r\n\r\n\r\n## Highlights\r\n\r\n- Migrate all the mmdet3d ops to mmcv\r\n- Support parallel waymo data converter\r\n- Add ScanNet instance segmentation dataset with metrics\r\n- Better compatibility for windows with CI support, op migration and bug fixes\r\n- Support loading annotations from Ceph\r\n\r\n## New Features\r\n\r\n- Add ScanNet instance segmentation dataset with metrics (#1230)\r\n- Support different random seeds for different ranks (#1321)\r\n- Support loading annotations from Ceph (#1325)\r\n- Support resuming from the latest checkpoint automatically (#1329)\r\n- Add windows CI (#1345)\r\n\r\n## Improvements\r\n\r\n- Update the table format and OpenMMLab project orders in [README.md](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fblob\u002Fmaster\u002FREADME.md) (#1272, #1283)\r\n- Migrate all the mmdet3d ops to mmcv (#1240, #1286, #1290, #1333)\r\n- Add `with_plane` flag in the KITTI data conversion (#1278)\r\n- Update instructions and links in the documentation (#1300, 1309, #1319)\r\n- Support parallel Waymo dataset converter and ground truth database generator (#1327)\r\n- Add quick installation commands to [getting_started.md](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fblob\u002Fmaster\u002Fdocs\u002Fen\u002Fgetting_started.md) (#1366)\r\n\r\n## Bug Fixes\r\n\r\n- Update nuimages configs to use new nms config style (#1258)\r\n- Fix the usage of np.long for windows compatibility (#1270)\r\n- Fix the incorrect indexing in `BasePoints` (#1274)\r\n- Fix the incorrect indexing in the [pillar_scatter.forward_single](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fblob\u002Fdev\u002Fmmdet3d\u002Fmodels\u002Fmiddle_encoders\u002Fpillar_scatter.py#L38) (#1280)\r\n- Fix unit tests that use GPUs (#1301)\r\n- Fix incorrect feature dimensions in `DynamicPillarFeatureNet` caused by previous upgrading of `PillarFeatureNet` (#1302)\r\n- Remove the `CameraPoints` constraint in `PointSample` (#1314)\r\n- Fix imprecise timestamps saving of Waymo dataset (#1327)\r\n\r\n## Contributors\r\n\r\nA total of 10 developers contributed to this release.\r\n\r\n@ZCMax, @ZwwWayne, @wHao-Wu, @Tai-Wang, @wangruohui, @zjwzcx, @Xiangxu-0103, @EdAyers, @hongye-dev, @zhanggefan \r\n\r\n## New Contributors\r\n* @VVsssssk made their first contribution in https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F1275\r\n* @Xiangxu-0103 made their first contribution in https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F1300\r\n* @Subjectivist made their first contribution in https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F1298\r\n* @EdAyers made their first contribution in https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F1258\r\n","2022-04-06T15:31:21",{"id":233,"version":234,"summary_zh":235,"released_at":236},136382,"v1.0.0rc0","### Compatibility\r\n\r\n- We refactor our three coordinate systems to make their rotation directions and origins more consistent, and further remove unnecessary hacks in different datasets and models. Therefore, please re-generate data information or convert the old version to the new one with our provided scripts. We will also provide updated checkpoints in the next version. Please refer to the [compatibility documentation](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fblob\u002Fv1.0.0.dev0\u002Fdocs\u002Fen\u002Fcompatibility.md) for more details.\r\n- Unify the camera keys for consistent transformation between coordinate systems on different datasets. The modification changes the key names to `lidar2img`, `depth2img`, `cam2img`, etc., for easier understanding. Customized codes using legacy keys may be influenced.\r\n- The next release will begin to move files of CUDA ops to [MMCV](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmcv). It will influence the way to import related functions. We will not break the compatibility but will raise a warning first and please prepare to migrate it.\r\n\r\n#### Highlights\r\n\r\n- Support new monocular 3D detectors: [PGD](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Ftree\u002Fv1.0.0.dev0\u002Fconfigs\u002Fpgd), [SMOKE](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Ftree\u002Fv1.0.0.dev0\u002Fconfigs\u002Fsmoke), [MonoFlex](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Ftree\u002Fv1.0.0.dev0\u002Fconfigs\u002Fmonoflex)\r\n- Support a new LiDAR-based detector: [PointRCNN](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Ftree\u002Fv1.0.0.dev0\u002Fconfigs\u002Fpoint_rcnn)\r\n- Support a new backbone: [DGCNN](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Ftree\u002Fv1.0.0.dev0\u002Fconfigs\u002Fdgcnn)\r\n- Support 3D object detection on the S3DIS dataset\r\n- Support compilation on Windows\r\n- Full benchmark for PAConv on S3DIS\r\n- Further enhancement for documentation, especially on the Chinese documentation\r\n\r\n#### New Features\r\n\r\n- Support 3D object detection on the S3DIS dataset (#835)\r\n- Support PointRCNN (#842, #843, #856, #974, #1022, #1109, #1125)\r\n- Support DGCNN (#896)\r\n- Support PGD (#938, #940, #948, #950, #964, #1014, #1065, #1070, #1157)\r\n- Support SMOKE (#939, #955, #959, #975, #988, #999, #1029)\r\n- Support MonoFlex (#1026, #1044, #1114, #1115, #1183)\r\n- Support CPU Training (#1196)\r\n\r\n#### Improvements\r\n\r\n- Support point sampling based on distance metric (#667, #840)\r\n- Refactor coordinate systems (#677, #774, #803, #899, #906, #912, #968, #1001)\r\n- Unify camera keys in PointFusion and transformations between different systems (#791, #805)\r\n- Refine documentation (#792, #827, #829, #836, #849, #854, #859, #1111, #1113, #1116, #1121, #1132, #1135, #1185, #1193, #1226)\r\n- Add a script to support benchmark regression (#808)\r\n- Benchmark PAConvCUDA on S3DIS (#847)\r\n- Support to download pdf and epub documentation (#850)\r\n- Change the `repeat` setting in Group-Free-3D configs to reduce training epochs (#855)\r\n- Support KITTI AP40 evaluation metric (#927)\r\n- Add the mmdet3d2torchserve tool for SECOND (#977)\r\n- Add code-spell pre-commit hook and fix typos (#995)\r\n- Support the latest numba version (#1043)\r\n- Set a default seed to use when the random seed is not specified (#1072)\r\n- Distribute mix-precision models to each algorithm folder (#1074)\r\n- Add abstract and a representative figure for each algorithm (#1086)\r\n- Upgrade pre-commit hook (#1088, #1217)\r\n- Support augmented data and ground truth visualization (#1092)\r\n- Add local yaw property for `CameraInstance3DBoxes` (#1130)\r\n- Lock the required numba version to 0.53.0 (#1159)\r\n- Support the usage of plane information for KITTI dataset (#1162)\r\n- Deprecate the support for \"python setup.py test\" (#1164)\r\n- Reduce the number of multi-process threads to accelerate training (#1168)\r\n- Support 3D flip augmentation for semantic segmentation (#1181)\r\n- Update README format for each model (#1195)\r\n\r\n### Bug Fixes\r\n\r\n- Fix compiling errors on Windows (#766)\r\n- Fix the deprecated nms setting in the ImVoteNet config (#828)\r\n- Use the latest `wrap_fp16_model` import from mmcv (#861)\r\n- Remove 2D annotations generation on Lyft (#867)\r\n- Update index files for the Chinese documentation to be consistent with the English version (#873)\r\n- Fix the nested list transpose in the CenterPoint head (#879)\r\n- Fix deprecated pretrained model loading for RegNet (#889)\r\n- Fix the incorrect dimension indices of rotations and testing config in the CenterPoint test time augmentation (#892)\r\n- Fix and improve visualization tools (#956, #1066, #1073)\r\n- Fix PointPillars FLOPs calculation error (#1075)\r\n- Fix missing dimension information in the SUN RGB-D data generation (#1120)\r\n- Fix incorrect anchor range settings in the PointPillars [config](https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fblob\u002Fmaster\u002Fconfigs\u002F_base_\u002Fmodels\u002Fhv_pointpillars_secfpn_kitti.py) for KITTI (#1163)\r\n- Fix incorrect model information in the RegNet metafile (#1184)\r\n- Fix bugs in non-distributed multi-gpu training and testing (#1197)\r\n- Fix a potential assertion error when generating corners from an empty box (#1212)\r","2022-03-01T06:51:54",{"id":238,"version":239,"summary_zh":240,"released_at":241},136383,"v0.18.1","### Improvements\r\n\r\n- Support Flip3D augmentation in semantic segmentation task (#1182)\r\n- Update regnet metafile (#1184)\r\n- Add point cloud annotation tools introduction in FAQ (#1185)\r\n- Add missing explanations of `cam_intrinsic` in the nuScenes dataset doc (#1193)\r\n\r\n### Bug Fixes\r\n\r\n- Deprecate the support for \"python setup.py test\" (#1164)\r\n- Fix the rotation matrix while rotation axis=0 (#1182)\r\n- Fix the bug in non-distributed multi-gpu training\u002Ftesting (#1197)\r\n- Fix a potential bug when generating corners of empty bounding boxes (#1212)\r\n\r\n#### Contributors\r\n\r\nA total of 4 developers contributed to this release.\r\n\r\n@ZwwWayne, @ZCMax, @Tai-Wang, @wHao-Wu\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fcompare\u002Fv0.18.0...v0.18.1","2022-02-09T11:28:23",{"id":243,"version":244,"summary_zh":245,"released_at":246},136384,"v0.18.0","### Highlights\r\n\r\n- Update the required minimum version of mmdet and mmseg\r\n\r\n### Improvements\r\n\r\n- Use the official markdownlint hook and add codespell hook for pre-committing (#1088)\r\n- Improve CI operation (#1095, #1102, #1103)\r\n- Use shared menu content from OpenMMLab's theme and remove duplicated contents from config (#1111)\r\n- Refactor the structure of documentation (#1113, #1121)\r\n- Update the required minimum version of mmdet and mmseg (#1147)\r\n\r\n### Bug Fixes\r\n\r\n- Fix symlink failure on Windows (#1096)\r\n- Fix the upper bound of mmcv version in the mminstall requirements (#1104)\r\n- Fix API documentation compilation and mmcv build errors (#1116)\r\n- Fix figure links and pdf documentation compilation (#1132, #1135)\r\n\r\n### Contributors\r\n\r\nA total of 4 developers contributed to this release.\r\n\r\n@ZwwWayne, @ZCMax, @Tai-Wang, @wHao-Wu\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fcompare\u002Fv0.17.3...v0.18.0","2022-01-05T16:26:40",{"id":248,"version":249,"summary_zh":250,"released_at":251},136385,"v0.17.3","## What's Changed\r\n* [Fix] Update mmcv version in dockerfile by @wHao-Wu in https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F1036\r\n* [Fix] Fix the memory-leak problem in init_detector by @Tai-Wang in https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F1045\r\n* [Fix] Fix default show value in show_result function and a typo in waymo_data_prep by @ZCMax in https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F1034\r\n* [Fix] Fix incorrect velo indexing when formatting boxes on nuScenes by @Tai-Wang in https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F1049\r\n* [Enhance] Clean unnecessary custom_imports in entrypoints by @ZCMax in https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F1068\r\n* [Doc] Add MMFlow into README by @ZCMax in https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F1067\r\n* Explicitly setting torch.cuda.device at init_model by @aldakata in https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F1056\r\n* [Fix] Fix PointPillars FLOPs calculation error for master branch by @ZCMax in https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F1076\r\n* [Enhance] Add mmFewShot in README by @ZCMax in https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F1085\r\n* Label visualization by @MilkClouds in https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F1050\r\n* [Enhance] add mmhuman3d in readme by @ZCMax in https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F1094\r\n* [Enhance] fix mmhuman3d reference by @ZCMax in https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F1100\r\n* Bump to v0.17.3 by @Tai-Wang in https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F1083\r\n\r\n## New Contributors\r\n* @aldakata made their first contribution in https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F1056\r\n* @MilkClouds made their first contribution in https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fpull\u002F1050\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fcompare\u002Fv0.17.2...v0.17.3","2021-12-06T14:31:52",{"id":253,"version":254,"summary_zh":255,"released_at":256},136386,"v0.17.2","### Improvements\r\n\r\n- Update Group-Free-3D and FCOS3D bibtex (#985)\r\n- Update the solutions for incompatibility of pycocotools in the FAQ (#993)\r\n- Add Chinese documentation for the KITTI (#1003) and Lyft (#1010) dataset tutorial\r\n- Add the H3DNet checkpoint converter for incompatible keys (#1007)\r\n\r\n### Bug Fixes\r\n\r\n- Update mmdetection and mmsegmentation version in the Dockerfile (#992)\r\n- Fix links in the Chinese documentation (#1015)\r\n\r\n### Contributors\r\n\r\nA total of 4 developers contributed to this release.\r\n\r\n@Tai-Wang, @wHao-Wu, @ZwwWayne, @ZCMax\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fopen-mmlab\u002Fmmdetection3d\u002Fcompare\u002Fv0.17.1...v0.17.2","2021-11-02T01:01:25"]