[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"similar-sipeed--MaixPy":3,"tool-sipeed--MaixPy":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",155373,2,"2026-04-14T11:34:08",[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。",108322,"2026-04-10T11:39:34",[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},6121,"gemini-cli","google-gemini\u002Fgemini-cli","gemini-cli 是一款由谷歌推出的开源 AI 命令行工具，它将强大的 Gemini 大模型能力直接集成到用户的终端环境中。对于习惯在命令行工作的开发者而言，它提供了一条从输入提示词到获取模型响应的最短路径，无需切换窗口即可享受智能辅助。\n\n这款工具主要解决了开发过程中频繁上下文切换的痛点，让用户能在熟悉的终端界面内直接完成代码理解、生成、调试以及自动化运维任务。无论是查询大型代码库、根据草图生成应用，还是执行复杂的 Git 操作，gemini-cli 都能通过自然语言指令高效处理。\n\n它特别适合广大软件工程师、DevOps 人员及技术研究人员使用。其核心亮点包括支持高达 100 万 token 的超长上下文窗口，具备出色的逻辑推理能力；内置 Google 搜索、文件操作及 Shell 命令执行等实用工具；更独特的是，它支持 MCP（模型上下文协议），允许用户灵活扩展自定义集成，连接如图像生成等外部能力。此外，个人谷歌账号即可享受免费的额度支持，且项目基于 Apache 2.0 协议完全开源，是提升终端工作效率的理想助手。",100752,"2026-04-10T01:20:03",[52,13,15,14],"插件",{"id":54,"name":55,"github_repo":56,"description_zh":57,"stars":58,"difficulty_score":32,"last_commit_at":59,"category_tags":60,"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],{"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":77,"owner_twitter":78,"owner_website":79,"owner_url":80,"languages":81,"stars":121,"forks":122,"last_commit_at":123,"license":124,"difficulty_score":125,"env_os":126,"env_gpu":127,"env_ram":128,"env_deps":129,"category_tags":137,"github_topics":138,"view_count":32,"oss_zip_url":76,"oss_zip_packed_at":76,"status":17,"created_at":147,"updated_at":148,"faqs":149,"releases":185},7506,"sipeed\u002FMaixPy","MaixPy","Easily create AI projects with Python on edge device","MaixPy 是一款专为边缘计算设备设计的开源框架，让开发者能够使用熟悉的 Python 语言轻松构建人工智能项目。它主要解决了在资源受限的嵌入式硬件上部署 AI 应用门槛高、开发流程复杂的问题，将复杂的底层硬件操作封装为简洁的接口。\n\n无论是物联网开发者、嵌入式工程师，还是希望快速验证原型的科研人员与创客，都能通过 MaixPy 高效地实现智能视觉识别、音频处理乃至大语言模型（LLM）等前沿功能。其核心亮点在于极致的代码效率，通常仅需十余行代码即可完成从摄像头采集到 AI 推理显示的全流程；同时支持串口等外设的直接调用，真正实现了“即插即用”。\n\n此外，MaixPy 不仅提供完善的文档和在线训练平台 MaixHub，还配套了图形化开发工具 MaixVision IDE，甚至提供了与 Python API 完全一致的 C\u002FC++ SDK，确保了从原型开发到量产部署的无缝衔接。如果你想在端侧设备上快速落地创意，MaixPy 是一个强大且友好的选择。","MaixPy (v4)\n======\n\n\u003Cdiv align=\"center\">\n\n![](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fsipeed_MaixPy_readme_a71a3139efc1.png)\n\n[![Back MaixCam 2 on Kickstarter](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FKickstarter-Back%20MaixCam%202-05CE78?style=for-the-badge&logo=kickstarter&logoColor=white)](https:\u002F\u002Fwww.kickstarter.com\u002Fprojects\u002Fzepan\u002Fmaixcam2-build-your-next-gen-4k-ai-camera)\n\n**🚀 [MaixCam 2 is LIVE: Build Your Next-Gen 4K AI Camera!](https:\u002F\u002Fwww.kickstarter.com\u002Fprojects\u002Fzepan\u002Fmaixcam2-build-your-next-gen-4k-ai-camera) 🚀**\n\n**MaixPy** (v4): Easily create AI projects with Python on edge device\n\n\u003Ch3>\n    \u003Ca href=\"https:\u002F\u002Fwiki.sipeed.com\u002Fmaixpy\u002Fdoc\u002Fen\u002Findex.html\"> Quick Start \u003C\u002Fa> |\n    \u003Ca href=\"https:\u002F\u002Fwiki.sipeed.com\u002Fmaixpy\u002Fen\u002Findex.html\"> Documentation \u003C\u002Fa> |\n    \u003Ca href=\"https:\u002F\u002Fwiki.sipeed.com\u002Fmaixpy\u002Fapi\u002Findex.html\"> API \u003C\u002Fa> |\n    \u003Ca href=\"https:\u002F\u002Fwiki.sipeed.com\u002Fmaixcam\"> Hardware \u003C\u002Fa>\n\u003C\u002Fh3>\n\n[![GitHub Repo stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fsipeed\u002FMaixPy?style=social)](https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixPy\u002Fstargazers)\n[![Apache 2.0](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Flicense-Apache%20v2.0-orange.svg)](\"https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixPy\u002Fblob\u002Fmain\u002FLICENSE.md)\n[![PyPI](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fv\u002Fmaixpy.svg)](https:\u002F\u002Fpypi.python.org\u002Fpypi\u002Fmaixpy\u002F)\n[![PyPI - Downloads](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fdm\u002Fmaixpy?label=pypi%20downloads)](https:\u002F\u002Fpypi.org\u002Fproject\u002Fmaixpy\u002F)\n![GitHub repo size](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Frepo-size\u002Fsipeed\u002Fmaixpy) \n[![Build MaixCAM](https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixPy\u002Factions\u002Fworkflows\u002Fbuild_maixcam.yml\u002Fbadge.svg)](https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixPy\u002Factions\u002Fworkflows\u002Fbuild_maixcam.yml)\n[![Trigger wiki](https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixPy\u002Factions\u002Fworkflows\u002Ftrigger_wiki.yml\u002Fbadge.svg)](https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixPy\u002Factions\u002Fworkflows\u002Ftrigger_wiki.yml)\n\nEnglish | [中文](.\u002FREADME_ZH.md)\n\n\u003C\u002Fdiv>\n\n![MaixCAM](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fsipeed_MaixPy_readme_3c47125fd146.png)\n\n\n## Feature Overview\n\nMaixPy offers simple Python programming combined with powerful edge computing hardware. Integrated hardware peripheral operations, video streaming, AI vision algorithms, audio algorithms, and LLM \u002F VLM etc. With its plug-and-play design, MaixPy enables you to quickly implement your intelligent projects.\n\nAdditionally, MaixPy provides the MaixVision IDE, MaixHub online training platform, detailed documentation, and even a C\u002FC++ SDK with identical APIs, ensuring seamless development and production deployment.\n\nHere is a partial video demonstration of the features. For more documentation, please visit the **official website: [wiki.sipeed.com\u002Fmaixpy\u002F](https:\u002F\u002Fwiki.sipeed.com\u002Fmaixpy\u002F)**\n\n> If you like this project, please click **Star** on the top right of the [MaixPy Project](https:\u002F\u002Fgithub.com\u002Fsipeed\u002Fmaixpy) to encourage us to develop more exciting content!\n\n[![](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fsipeed_MaixPy_readme_ed2f95e02bc9.jpg)](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=qV1lw0UVUYI)\n\n\n## Concise and Efficient Code (API) Design\n\nWith MaixPy you can easily create **AI vision project** within **10 lines of code**:\n\n```python\nfrom maix import camera, display, image, nn\n\nclassifier = nn.Classifier(model=\"\u002Froot\u002Fmodels\u002Fmobilenetv2.mud\")\ncam = camera.Camera(classifier.input_width(), classifier.input_height(), classifier.input_format())\ndisp = display.Display()\n\nwhile 1:\n    img = cam.read()\n    res = classifier.classify(img)\n    max_idx, max_prob = res[0]\n    msg = f\"{max_prob:5.2f}: {classifier.labels[max_idx]}\"\n    img.draw_string(10, 10, msg, image.COLOR_RED)\n    disp.show(img)\n```\n\nResult:\n![](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fsipeed_MaixPy_readme_f5cf1b4b1594.gif)\n\n## Edge(embeded) friendly\n\nSimply use hardware **peripheral** like serial port:\n\n```python\nfrom maix import uart\n\ndevices = uart.list_devices()\n\nserial = uart.UART(devices[0], 115200)\nserial.write_str(\"hello world\")\nprint(\"received:\", serial.read(timeout = 2000))\n```\n\n## MaixVision workstation\n\nWe also provide a handy **[MaixVision](https:\u002F\u002Fwiki.sipeed.com\u002Fen\u002Fmaixvision)** workstation software to make development easier and faster:\n\n\u003Cvideo playsinline controls muted preload src=\"https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixPy\u002Fassets\u002F8625829\u002F1168a282-d7c2-45bc-9ffb-c00de1ca24f5\" type=\"video\u002Fmp4\">\nMaixVision\n\u003C\u002Fvideo>\n\n## MaixHub online platform\n\n**[MaixHub](https:\u002F\u002Fmaixhub.com)** provide free online AI train service, one click to train AI model and deploy to MaixCAM even you have no AI knowledge and expensive training equipment.\n\n![MaixHub](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fsipeed_MaixPy_readme_503dd201baf5.jpg)\n\n\n## Hardware platform MaixCAM\n\nAnd we provide two powerful hardware platform **[MaixCAM2](https:\u002F\u002Fwiki.sipeed.com\u002Fmaixcam2)**, **[MaixCAM](https:\u002F\u002Fwiki.sipeed.com\u002Fmaixcam)** and **[MaixCAM-Pro](https:\u002F\u002Fwiki.sipeed.com\u002Fmaixcam-pro)**.\n\n![MaixCAM](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fsipeed_MaixPy_readme_3c47125fd146.png)\n\n\n## Performance comparison\n\nK210 and v831 are outdated, they have many limitations in memory, performance, NPU operators missing etc.\u003Cbr>\nNo matter you are using them or new comer, it's recommended to upgrade to MaixCAM and MaixPy v4.\u003Cbr>\n\nHere's the comparison between them:\n\n| Feature      | Maix-I K210 | MaixCAM | MaixCAM2 |\n| --------- | ----------- | ------- | ------- |\n| CPU       | 400MHz RISC-V x2 |  **1GHz RISC-V(Linux)\u003Cbr>700MHz RISC-V(RTOS)\u003Cbr>25~300MHz 8051(Low Power)** | \u003Cspan class=\"strong2\">1.2GHz A53 x2(Linux)\u003C\u002Fspan>\u003Cbr>RISC-V 32bit E907(RTT) |\n| Memory      | 6MB SRAM         | **256MB DDR3** | \u003Cspan class=\"strong2\">1GB \u002F 4GB LPDDR4\u003C\u002Fspan> |\n| NPU       | 0.25Tops@INT8\u003Cbr>official says 1T but... | **1Tops@INT8** | \u003Cspan class=\"strong2\">3.2Tops@INT8\u003C\u002Fspan> |\n| Encoder   | ❌               | **2880x1620@30fps H.254\u002FH.265\u002FJPEG** | \u003Cspan class=\"strong2\">3840*2160@30fps H.254\u002FH.265\u002FJPEG\u003C\u002Fspan> |\n| Decoder   | ❌               | **2880x1620@30fps H.264\u002FJPEG** | **1080p@60fps H.264\u002FJPEG** |\n| Screen      | 2.4\" 320x240     | **2.3\" 552x368**(MaixCAM)\u003Cbr\u002F>**2.4\" 640x480**(MaixCAM-Pro)\u003Cbr\u002F>5\" 1280x720\u003Cbr\u002F>7\" 1280x800\u003Cbr\u002F>10“ 1280x800| **2.4\" 640x480**\u003Cbr\u002F>5\" 1280x720\u003Cbr\u002F>7\" 1280x800\u003Cbr\u002F>10“ 1280x800 |\n| Touchscree    | ❌               | **2.3\" 552x368**\u002F**2.4\" 640x480** | **2.4\" 640x480** |\n| Camera    | 30W              | **500W(5M)** | \u003Cspan class=\"strong2\">800W(8M)\u003C\u002Fspan> |\n| AI ISP    | ❌               | ❌           | \u003Cspan class=\"strong2\">✅\u003C\u002Fspan> |\n| WiFi      | 2.4G             | **WiFi6** 2.4G\u002F5G | **WiFi6** 2.4G\u002F5G |\n| BLE       | ❌               | **BLE5.4** | **BLE5.4** |\n| USB       | ❌               | **USB2.0** | **USB2.0** |\n| Ethernet    | ❌               | 100M(Optional) | \u003Cspan class=\"strong2\">100M(on board FPC, can convert to RJ45 module)\u003C\u002Fspan> |\n| SD Card | SPI              | **SDIO** | **SDIO** |\n| OS  | RTOS             | **Linux(BuildRoot) + RTOS** | Linux(\u003Cspan class=\"strong2\">Ubuntu\u003C\u002Fspan>) + RTT |\n| Porgraming Language  | C \u002F C++ \u002F MicroPython | C \u002F **C++ \u002F Python3** | C \u002F **C++ \u002F Python3** |\n| SDK    | MaixPy-v1             | **MaixCDK + MaixPy v4\u003Cbr>+ opencv + numpy + ...**\u003Cbr>Pure Python package or cross-compile manually | **MaixCDK + MaixPy v4\u003Cbr>+ opencv + numpy + scipy + ...**\u003Cbr>\u003Cspan class=\"strong2\">Many AArch64 pre-compiled packages, and support compile on board\u003C\u002Fspan> |\n| PC Software   | MaixPy IDE            | **MaixVision** Workstation | **MaixVision** Workstation |\n| Documentation             | ⭐️⭐️⭐️⭐️     |  🌟🌟🌟🌟🌟 | 🌟🌟🌟🌟🌟 |\n| Online AI train       | ⭐️⭐️⭐️        |  🌟🌟🌟🌟🌟 | 🌟🌟🌟🌟🌟 |\n| Official APPs         | ⭐️             |  🌟🌟🌟🌟🌟 | 🌟🌟🌟🌟🌟 |\n| Ease of use           | ⭐️⭐️⭐️⭐️      |  🌟🌟🌟🌟🌟 | 🌟🌟🌟🌟🌟 |\n| AI classify(224x224) | MobileNetv1 50fps\u003Cbr>MobileNetv2 ❌\u003Cbr>Resnet ❌ | MobileNetv2 **130fps**\u003Cbr>Resnet18 **62fps**\u003Cbr>Resnet50 **28fps** | MobileNetv2 \u003Cspan class=\"strong2\">1218fps\u003C\u002Fspan>\u003Cbr>Resnet50 \u003Cspan class=\"strong2\">200fps\u003C\u002Fspan> |\n| AI detect\u003Cdiv class=\"comment\">only forward part \u002F\u003Cbr>\\[include pre-post process parts(Python)\\] \u002F\u003Cbr>\\[dual buff mode(Python)\\]\u003C\u002Fdiv> | \u003Cdiv class=\"main_items\">**YOLOv2**:\u003Cdiv class=\"sub_items\">224x224: 15fps\u003C\u002Fdiv>\u003C\u002Fdiv> |  \u003Cdiv class=\"main_items\">**YOLOv5s**:\u003Cdiv class=\"sub_items\">224x224: **100fps**\u003Cbr>320x256 **70fps**\u003Cbr>640x640: **15fps**\u003C\u002Fdiv>\u003C\u002Fdiv>       \u003Cdiv class=\"main_items\">**YOLOv8n**:\u003Cdiv class=\"sub_items\">640x640: **23fps**\u003C\u002Fdiv>\u003C\u002Fdiv>      \u003Cdiv class=\"main_items\">**YOLO11n**:\u003Cdiv class=\"sub_items\">224x224: **175fps**\u003Cbr>320x224: **120fps**\u003Cbr>320x320: **95fps**\u003Cbr>640x640: **23fps**\u003C\u002Fdiv>\u003C\u002Fdiv>                |                \u003Cdiv class=\"main_items\">**YOLOv5s**:\u003Cdiv class=\"sub_items\">224x224: \u003Cspan class=\"strong2\">495fps\u003C\u002Fspan>\u003Cbr>320x256: \u003Cspan class=\"strong2\">400fps\u003C\u002Fspan>\u003Cbr>640x480: \u003Cspan class=\"strong2\">106fps \u002F 73fps \u002F 103fps\u003C\u002Fspan>\u003Cbr>640x640: \u003Cspan class=\"strong2\">80fps\u003C\u002Fspan>\u003C\u002Fdiv>\u003C\u002Fdiv>                \u003Cdiv class=\"main_items\">**YOLO11n**:\u003Cdiv class=\"sub_items\">224x224: \u003Cspan class=\"strong2\">1214fps\u003C\u002Fspan>\u003Cbr>640x480: \u003Cspan class=\"strong2\">168fps \u002F 77fps \u002F 143fps\u003C\u002Fspan>\u003Cbr>640x640: \u003Cspan class=\"strong2\">113fps \u002F 56fps \u002F 98fps\u003C\u002Fspan>\u003C\u002Fdiv>\u003C\u002Fdiv>    \u003Cdiv class=\"main_items\">**YOLO11s**:\u003Cdiv class=\"sub_items\">640x480: \u003Cspan class=\"strong2\">87fps \u002F 53fps \u002F 83fps\u003C\u002Fspan>\u003Cbr>640x640: \u003Cspan class=\"strong2\">62fps \u002F 39fps \u002F 59fps\u003C\u002Fspan>\u003C\u002Fdiv>\u003C\u002Fdiv>   \u003Cdiv class=\"main_items\">**YOLO11l**:\u003Cdiv class=\"sub_items\">640x640: \u003Cspan class=\"strong2\">19fps \u002F 16fps \u002F 19fps\u003C\u002Fspan>\u003C\u002Fdiv>\u003C\u002Fdiv>                     |\n| LLM           | ❌              |  ❌           |  \u003Cspan class=\"strong2\">Qwen\u002FDeepSeek 0.5B(fftf: 640ms, 9 tokens\u002Fs)\u003Cbr>Qwen\u002FDeepSeek 1.5B(fftf: 1610ms, 4 tokens\u002Fs) \u003Cbr> VLM(InterVL 1B) \u003Cbr>Mode models\u003C\u002Fspan> |\n| OpenMV algorithms |  \u003Cdiv class=\"comment\">test image refer to \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixPy\u002Ftree\u002Fmain\u002Fprojects\u002Fapp_benchmark\">Benchmark APP\u003C\u002Fa>  |                    \u003Cdiv class=\"comment\">test image refer to \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixPy\u002Ftree\u002Fmain\u002Fprojects\u002Fapp_benchmark\">Benchmark APP\u003C\u002Fa>\u003Cbr>test date: 2025.8.22，update may have optimization\u003C\u002Fdiv>              |       \u003Cdiv class=\"comment\">test image refer to \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixPy\u002Ftree\u002Fmain\u002Fprojects\u002Fapp_benchmark\">Benchmark APP\u003C\u002Fa>\u003Cbr>test date: 2025.8.22，update may have optimization\u003C\u002Fdiv>                             |\n|   \u003Cdiv class=\"right second\">Binary\u003C\u002Fdiv>  | Gray 320x240: 7.4ms (135fps)\u003Cbr>Gray 640x480: ❌\u003Cbr>RGB 320x240: 11.3ms (88.5fps)\u003Cbr>RGB 640x480: ❌ | Gray 320x240: **3.1ms (326fps)**\u003Cbr>Gray 640x480: **11ms (90fps)**\u003Cbr>RGB 320x240: **13.2ms (75fps)**\u003Cbr>RGB 640x480: **52.8ms (18fps)**        | Gray 320x240: \u003Cspan class=\"strong2\">1.3ms (799fps)\u003C\u002Fspan> \u003Cbr>Gray 640x480: \u003Cspan class=\"strong2\">4.8ms (206fps)\u003C\u002Fspan>\u003Cbr>RGB 320x240: \u003Cspan class=\"strong2\">3.4ms (294fps)\u003C\u002Fspan>\u003Cbr>RGB 640x480: \u003Cspan class=\"strong2\">13.3ms (75fps)\u003C\u002Fspan> |\n|   \u003Cdiv class=\"right second\">Find blobs\u003C\u002Fdiv>        | 320x240: 8.8ms (114fps) \u003Cbr>640x480: ❌| 320x240: **7ms (143fps)**  \u003Cbr>640x480: **20ms (50fps)**         | 320x240: \u003Cspan class=\"strong2\">3.7ms (271fps)\u003C\u002Fspan>\u003Cbr>640x480: \u003Cspan class=\"strong2\">11.1ms (89fps)\u003C\u002Fspan>  |\n|   \u003Cdiv class=\"right second\">1channel histogram\u003C\u002Fdiv>  | 320x240: **7.7ms (130fps)**\u003Cbr>640x480: ❌ | 320x240: **10.9ms (91fps)**\u003Cbr>640x480: **42.8ms (23fps)**       | 320x240: \u003Cspan class=\"strong2\">1.5ms (661fps)\u003C\u002Fspan>\u003Cbr>640x480: \u003Cspan class=\"strong2\">5.9ms (168fps)\u003C\u002Fspan>    |\n|   \u003Cdiv class=\"right second\">QR Code\u003C\u002Fdiv>        | 320x240: **130.8ms (7.6fps)** \u003Cbr>640x480: ❌| 640x480: 136.9ms (7fps)\u003Cbr>NPU acceleration：\u003Cbr>  320x240: **22.1ms (45fps)**\u003Cbr>  640x480: 57.6ms (17fps)  | 640x480: 57.9ms (17fps)\u003Cbr>NPU acceleration：\u003Cbr>  320x240: \u003Cspan class=\"strong2\">9.2ms (109fps)\u003C\u002Fspan>   \u003Cbr>  640x480: \u003Cspan class=\"strong2\">23.2ms (43fps)\u003C\u002Fspan> |\n| OpenCV algorithms     |   | \u003Cdiv class=\"comment\">test image refer to \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixPy\u002Ftree\u002Fmain\u002Fprojects\u002Fapp_benchmark\">Benchmark APP\u003C\u002Fa>\u003Cbr>test date: 2025.8.22，update may have optimization\u003C\u002Fdiv>    | \u003Cdiv class=\"comment\">test image refer to \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixPy\u002Ftree\u002Fmain\u002Fprojects\u002Fapp_benchmark\">Benchmark APP\u003C\u002Fa>\u003Cbr>test date: 2025.8.22，update may have optimization\u003C\u002Fdiv>  |\n|   \u003Cdiv class=\"right second\">Binary\u003C\u002Fdiv>             | ❌  | Gray 320x240: **2.2ms (463fps)**     \u003Cbr>Gray 640x480: **7.1ms (140fps)** | Gray 320x240: \u003Cspan class=\"strong2\">0.1ms (8174fps)\u003C\u002Fspan>  \u003Cbr>Gray 640x480: \u003Cspan class=\"strong2\">0.3ms (2959fps)\u003C\u002Fspan>  |\n|   \u003Cdiv class=\"right second\">Gray adaptive binary\u003C\u002Fdiv> | ❌  | 320x240: **5.8ms (171fps)**     \u003Cbr>640x480: **21.3ms (46fps)**  | 320x240: \u003Cspan class=\"strong2\">1.6ms (608fps)\u003C\u002Fspan>  \u003Cbr>640x480: \u003Cspan class=\"strong2\">6.3ms (159fps)\u003C\u002Fspan> |\n|   \u003Cdiv class=\"right second\">1channel histogram\u003C\u002Fdiv>       | ❌  | 320x240: **1ms (1000fps)**     \u003Cbr>640x480: **6.2ms (160fps)**   | 320x240: \u003Cspan class=\"strong2\">0.4ms (2308fps)\u003C\u002Fspan>  \u003Cbr>640x480: \u003Cspan class=\"strong2\">1.7ms (604fps)\u003C\u002Fspan>  |\n|   \u003Cdiv class=\"right second\">Find Contours\u003C\u002Fdiv>           | ❌  | 320x240: **2.8ms (351fps)**    \u003Cbr>640x480: **8.6ms (116fps)**   | 320x240: \u003Cspan class=\"strong2\">0.4ms (2286fps)\u003C\u002Fspan>  \u003Cbr>640x480: \u003Cspan class=\"strong2\">1.4ms (692fps)\u003C\u002Fspan>  |\n\n## Maix Ecosystem\n\nMaixPy not only a Python SDK, but have a whole ecosystem, including hardware, software, tools, docs, even cloud platform etc.\nSee the picture below:\n\n![](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fsipeed_MaixPy_readme_df583d1f54a4.png)\n\n\n## Who are using MaixPy?\n\n* **AI Algorithm Engineer** who want to deploy your AI model to embedded devices.\n> MaixPy provide easy-to-use API to access NPU, and docs to help you develop your AI model.\n* **STEM** teacher who wants to teach AI and embedded devices to students.\n> MaixPy provide easy-to-use API, PC tools, online AI train service ... Let you focus on teaching AI, not the hardware and complicated software usage.\n* **Maker** who want to make some cool projects but don't want to learn complicated hardware and software.\n> MaixPy provide Python API, so all you need is learn basic Python syntax, and MaixPy's API is so easy to use, you can make your project even in a few minutes.\n* **Engineer** who want to make some projects but want a prototype as soon as possible.\n> MaixPy is easy to build projects, and provide corresponding C++ SDK, so you can directly use MaixPy to deploy or transfer Python code to C++ in a few minutes.\n* **Students** who want to learn AI, embedded development.\n> We provide many docs and tutorials, and lot of open source code, to help you find learning route, and grow up step by step. From simple Python programming to `Vision`, `AI`, `Audio`, `Linux`, `RTOS` etc.\n* **Enterprise** who want to develop AI vision products but have no time or engineers to develop complicated embedded system.\n> Use MaixPy even graphic programming to develop your products with no more employees and time. For example, add a AI QA system to your production line, or add a AI security monitor to your office as your demand.\n* **Contestants** who want to win the competition.\n> MaixPy integrate many functions and easy to use, fasten your work to win the competition in limited time. There are already many contestants win the competition with MaixPy.\n\n\n## Compile Source Code\n\nIf you want to compile MaixPy firmware from source code, refer to [Build MaixPy source code](https:\u002F\u002Fwiki.sipeed.com\u002Fmaixpy\u002Fdoc\u002Fen\u002Fsource_code\u002Fbuild.html) page.\n\n\n## License\n\nAll files in this repository are under the terms of the [Apache License 2.0 Sipeed Ltd.](.\u002FLICENSE) except the third-party libraries or have their own license.\n\n\n## Community\n\n* Project sharing: [maixhub.com\u002Fshare](https:\u002F\u002Fmaixhub.com\u002Fshare)\n* App sharing: [maixhub.com\u002Fapp](https:\u002F\u002Fmaixhub.com\u002Fapp)\n* Discussion: [maixhub.com\u002Fdiscussion](https:\u002F\u002Fmaixhub.com\u002Fdiscussion)\n* QQ group: 862340358\n* Telegram: [t.me\u002Fmaixpy](https:\u002F\u002Ft.me\u002Fmaixpy)\n* Github issues: [github.com\u002Fsipeed\u002Fmaixpy\u002Fissues](https:\u002F\u002Fgithub.com\u002Fsipeed\u002Fmaixpy\u002Fissues)\n\n\n\n","MaixPy (v4)\n======\n\n\u003Cdiv align=\"center\">\n\n![](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fsipeed_MaixPy_readme_a71a3139efc1.png)\n\n[![在Kickstarter上支持MaixCam 2](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FKickstarter-Back%20MaixCam%202-05CE78?style=for-the-badge&logo=kickstarter&logoColor=white)](https:\u002F\u002Fwww.kickstarter.com\u002Fprojects\u002Fzepan\u002Fmaixcam2-build-your-next-gen-4k-ai-camera)\n\n**🚀 [MaixCam 2已上线：打造你的下一代4K AI相机！](https:\u002F\u002Fwww.kickstarter.com\u002Fprojects\u002Fzepan\u002Fmaixcam2-build-your-next-gen-4k-ai-camera) 🚀**\n\n**MaixPy** (v4)：在边缘设备上用Python轻松创建AI项目\n\n\u003Ch3>\n    \u003Ca href=\"https:\u002F\u002Fwiki.sipeed.com\u002Fmaixpy\u002Fdoc\u002Fen\u002Findex.html\"> 快速入门 \u003C\u002Fa> |\n    \u003Ca href=\"https:\u002F\u002Fwiki.sipeed.com\u002Fmaixpy\u002Fen\u002Findex.html\"> 文档 \u003C\u002Fa> |\n    \u003Ca href=\"https:\u002F\u002Fwiki.sipeed.com\u002Fmaixpy\u002Fapi\u002Findex.html\"> API \u003C\u002Fa> |\n    \u003Ca href=\"https:\u002F\u002Fwiki.sipeed.com\u002Fmaixcam\"> 硬件 \u003C\u002Fa>\n\u003C\u002Fh3>\n\n[![GitHub仓库星标数](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fsipeed\u002FMaixPy?style=social)](https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixPy\u002Fstargazers)\n[![Apache 2.0](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Flicense-Apache%20v2.0-orange.svg)](\"https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixPy\u002Fblob\u002Fmain\u002FLICENSE.md\")\n[![PyPI](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fv\u002Fmaixpy.svg)](https:\u002F\u002Fpypi.python.org\u002Fpypi\u002Fmaixpy\u002F)\n[![PyPI - 下载量](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fdm\u002Fmaixpy?label=pypi%20downloads)](https:\u002F\u002Fpypi.org\u002Fproject\u002Fmaixpy\u002F)\n![GitHub仓库大小](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Frepo-size\u002Fsipeed\u002Fmaixpy) \n[![构建MaixCAM](https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixPy\u002Factions\u002Fworkflows\u002Fbuild_maixcam.yml\u002Fbadge.svg)](https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixPy\u002Factions\u002Fworkflows\u002Fbuild_maixcam.yml)\n[![触发维基更新](https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixPy\u002Factions\u002Fworkflows\u002Ftrigger_wiki.yml\u002Fbadge.svg)](https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixPy\u002Factions\u002Fworkflows\u002Ftrigger_wiki.yml)\n\nEnglish | [中文](.\u002FREADME_ZH.md)\n\n\u003C\u002Fdiv>\n\n![MaixCAM](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fsipeed_MaixPy_readme_3c47125fd146.png)\n\n\n## 功能概览\n\nMaixPy提供简洁的Python编程接口，结合强大的边缘计算硬件。它集成了硬件外设操作、视频流处理、AI视觉算法、音频算法以及LLM\u002FVLM等功能。凭借即插即用的设计，MaixPy使您能够快速实现智能项目。\n\n此外，MaixPy还提供了MaixVision IDE、MaixHub在线训练平台、详尽的文档，甚至拥有API完全一致的C\u002FC++ SDK，确保开发与生产部署的无缝衔接。\n\n以下是部分功能的视频演示。更多文档请访问**官方网站：[wiki.sipeed.com\u002Fmaixpy\u002F](https:\u002F\u002Fwiki.sipeed.com\u002Fmaixpy\u002F)**\n\n> 如果您喜欢这个项目，请点击[MaixPy项目](https:\u002F\u002Fgithub.com\u002Fsipeed\u002Fmaixpy)页面右上角的“Star”按钮，以鼓励我们开发更多精彩内容！\n\n[![](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fsipeed_MaixPy_readme_ed2f95e02bc9.jpg)](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=qV1lw0UVUYI)\n\n\n## 简洁高效的代码（API）设计\n\n使用MaixPy，您只需**10行代码**即可轻松创建一个**AI视觉项目**：\n\n```python\nfrom maix import camera, display, image, nn\n\nclassifier = nn.Classifier(model=\"\u002Froot\u002Fmodels\u002Fmobilenetv2.mud\")\ncam = camera.Camera(classifier.input_width(), classifier.input_height(), classifier.input_format())\ndisp = display.Display()\n\nwhile 1:\n    img = cam.read()\n    res = classifier.classify(img)\n    max_idx, max_prob = res[0]\n    msg = f\"{max_prob:5.2f}: {classifier.labels[max_idx]}\"\n    img.draw_string(10, 10, msg, image.COLOR_RED)\n    disp.show(img)\n```\n\n结果：\n![](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fsipeed_MaixPy_readme_f5cf1b4b1594.gif)\n\n## 边缘（嵌入式）友好\n\n只需使用串口等硬件**外设**即可：\n\n```python\nfrom maix import uart\n\ndevices = uart.list_devices()\n\nserial = uart.UART(devices[0], 115200)\nserial.write_str(\"hello world\")\nprint(\"received:\", serial.read(timeout = 2000))\n```\n\n## MaixVision工作站\n\n我们还提供便捷的**[MaixVision](https:\u002F\u002Fwiki.sipeed.com\u002Fen\u002Fmaixvision)**工作站软件，让开发更加简单快捷：\n\n\u003Cvideo playsinline controls muted preload src=\"https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixPy\u002Fassets\u002F8625829\u002F1168a282-d7c2-45bc-9ffb-c00de1ca24f5\" type=\"video\u002Fmp4\">\nMaixVision\n\u003C\u002Fvideo>\n\n## MaixHub在线平台\n\n**[MaixHub](https:\u002F\u002Fmaixhub.com)** 提供免费的在线AI训练服务，无需任何AI知识和昂贵的训练设备，即可一键训练AI模型并部署到MaixCAM上。\n\n![MaixHub](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fsipeed_MaixPy_readme_503dd201baf5.jpg)\n\n\n## 硬件平台MaixCAM\n\n我们还提供了两款强大的硬件平台**[MaixCAM2](https:\u002F\u002Fwiki.sipeed.com\u002Fmaixcam2)**、**[MaixCAM](https:\u002F\u002Fwiki.sipeed.com\u002Fmaixcam)**和**[MaixCAM-Pro](https:\u002F\u002Fwiki.sipeed.com\u002Fmaixcam-pro)**。\n\n![MaixCAM](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fsipeed_MaixPy_readme_3c47125fd146.png)\n\n\n## 性能对比\n\nK210和v831已经过时，在内存、性能以及NPU算子缺失等方面存在诸多限制。\u003Cbr>\n无论您是仍在使用它们，还是刚刚入门，都建议升级到MaixCAM和MaixPy v4。\u003Cbr>\n\n以下是它们之间的对比：\n\n| 特性      | Maix-I K210 | MaixCAM | MaixCAM2 |\n| --------- | ----------- | ------- | ------- |\n| CPU       | 400MHz RISC-V x2 | **1GHz RISC-V(Linux)\u003Cbr>700MHz RISC-V(RTOS)\u003Cbr>25~300MHz 8051(Low Power)** | \u003Cspan class=\"strong2\">1.2GHz A53 x2(Linux)\u003C\u002Fspan>\u003Cbr>RISC-V 32bit E907(RTT) |\n| 内存      | 6MB SRAM         | **256MB DDR3** | \u003Cspan class=\"strong2\">1GB \u002F 4GB LPDDR4\u003C\u002Fspan> |\n| NPU       | 0.25Tops@INT8\u003Cbr>官方称1T但… | **1Tops@INT8** | \u003Cspan class=\"strong2\">3.2Tops@INT8\u003C\u002Fspan> |\n| 编码器   | ❌               | **2880x1620@30fps H.254\u002FH.265\u002FJPEG** | \u003Cspan class=\"strong2\">3840*2160@30fps H.254\u002FH.265\u002FJPEG\u003C\u002Fspan> |\n| 解码器   | ❌               | **2880x1620@30fps H.264\u002FJPEG** | **1080p@60fps H.264\u002FJPEG** |\n| 屏幕      | 2.4\" 320x240     | **2.3\" 552x368**(MaixCAM)\u003Cbr\u002F>**2.4\" 640x480**(MaixCAM-Pro)\u003Cbr\u002F>5\" 1280x720\u003Cbr\u002F>7\" 1280x800\u003Cbr\u002F>10“ 1280x800| **2.4\" 640x480**\u003Cbr\u002F>5\" 1280x720\u003Cbr\u002F>7\" 1280x800\u003Cbr\u002F>10“ 1280x800 |\n| 触摸屏    | ❌               | **2.3\" 552x368**\u002F**2.4\" 640x480** | **2.4\" 640x480** |\n| 摄像头    | 30W              | **500W(5M)** | \u003Cspan class=\"strong2\">800W(8M)\u003C\u002Fspan> |\n| AI ISP    | ❌               | ❌           | \u003Cspan class=\"strong2\">✅\u003C\u002Fspan> |\n| WiFi      | 2.4G             | **WiFi6** 2.4G\u002F5G | **WiFi6** 2.4G\u002F5G |\n| BLE       | ❌               | **BLE5.4** | **BLE5.4** |\n| USB       | ❌               | **USB2.0** | **USB2.0** |\n| 以太网    | ❌               | 100M(Optional) | \u003Cspan class=\"strong2\">100M(on board FPC, can convert to RJ45 module)\u003C\u002Fspan> |\n| SD卡 | SPI              | **SDIO** | **SDIO** |\n| 操作系统  | RTOS             | **Linux(BuildRoot) + RTOS** | Linux(\u003Cspan class=\"strong2\">Ubuntu\u003C\u002Fspan>) + RTT |\n| 编程语言  | C \u002F C++ \u002F MicroPython | C \u002F **C++ \u002F Python3** | C \u002F **C++ \u002F Python3** |\n| SDK    | MaixPy-v1             | **MaixCDK + MaixPy v4\u003Cbr>+ opencv + numpy + ...**\u003Cbr>Pure Python package or cross-compile manually | **MaixCDK + MaixPy v4\u003Cbr>+ opencv + numpy + scipy + ...**\u003Cbr>\u003Cspan class=\"strong2\">Many AArch64 pre-compiled packages, and support compile on board\u003C\u002Fspan> |\n| PC软件   | MaixPy IDE            | **MaixVision** Workstation | **MaixVision** Workstation |\n| 文档             | ⭐️⭐️⭐️⭐️     | 🌟🌟🌟🌟🌟 | 🌟🌟🌟🌟🌟 |\n| 在线AI训练       | ⭐️⭐️⭐️        | 🌟🌟🌟🌟🌟 | 🌟🌟🌟🌟🌟 |\n| 官方APPs         | ⭐️             | 🌟🌟🌟🌟🌟 | 🌟🌟🌟🌟🌟 |\n| 易用性           | ⭐️⭐️⭐️⭐️      | 🌟🌟🌟🌟🌟 | 🌟🌟🌟🌟🌟 |\n| AI分类(224x224) | MobileNetv1 50fps\u003Cbr>MobileNetv2 ❌\u003Cbr>Resnet ❌ | MobileNetv2 **130fps**\u003Cbr>Resnet18 **62fps**\u003Cbr>Resnet50 **28fps** | MobileNetv2 \u003Cspan class=\"strong2\">1218fps\u003C\u002Fspan>\u003Cbr>Resnet50 \u003Cspan class=\"strong2\">200fps\u003C\u002Fspan> |\n| AI检测\u003Cdiv class=\"comment\">仅前向部分 \u002F\u003Cbr>\\[包含预后处理(Python)\\] \u002F\u003Cbr>\\[双缓冲模式(Python)\\]\u003C\u002Fdiv> | \u003Cdiv class=\"main_items\">**YOLOv2**:\u003Cdiv class=\"sub_items\">224x224: 15fps\u003C\u002Fdiv>\u003C\u002Fdiv> |  \u003Cdiv class=\"main_items\">**YOLOv5s**:\u003Cdiv class=\"sub_items\">224x224: **100fps**\u003Cbr>320x256 **70fps**\u003Cbr>640x640: **15fps**\u003C\u002Fdiv>\u003C\u002Fdiv>       \u003Cdiv class=\"main_items\">**YOLOv8n**:\u003Cdiv class=\"sub_items\">640x640: **23fps**\u003C\u002Fdiv>\u003C\u002Fdiv>      \u003Cdiv class=\"main_items\">**YOLO11n**:\u003Cdiv class=\"sub_items\">224x224: **175fps**\u003Cbr>320x224: **120fps**\u003Cbr>320x320: **95fps**\u003Cbr>640x640: **23fps**\u003C\u002Fdiv>\u003C\u002Fdiv>                |                \u003Cdiv class=\"main_items\">**YOLOv5s**:\u003Cdiv class=\"sub_items\">224x224: \u003Cspan class=\"strong2\">495fps\u003C\u002Fspan>\u003Cbr>320x256: \u003Cspan class=\"strong2\">400fps\u003C\u002Fspan>\u003Cbr>640x640: \u003Cspan class=\"strong2\">80fps\u003C\u002Fspan>\u003C\u002Fdiv>\u003C\u002Fdiv>                \u003Cdiv class=\"main_items\">**YOLO11n**:\u003Cdiv class=\"sub_items\">224x224: \u003Cspan class=\"strong2\">1214fps\u003C\u002Fspan>\u003Cbr>640x480: \u003Cspan class=\"strong2\">168fps \u002F 77fps \u002F 143fps\u003C\u002Fspan>\u003Cbr>640x640: \u003Cspan class=\"strong2\">113fps \u002F 56fps \u002F 98fps\u003C\u002Fspan>\u003C\u002Fdiv>\u003C\u002Fdiv>    \u003Cdiv class=\"main_items\">**YOLO11s**:\u003Cdiv class=\"sub_items\">640x480: \u003Cspan class=\"strong2\">87fps \u002F 53fps \u002F 83fps\u003C\u002Fspan>\u003Cbr>640x640: \u003Cspan class=\"strong2\">62fps \u002F 39fps \u002F 59fps\u003C\u002Fspan>\u003C\u002Fdiv>\u003C\u002Fdiv>   \u003Cdiv class=\"main_items\">**YOLO11l**:\u003Cdiv class=\"sub_items\">640x640: \u003Cspan class=\"strong2\">19fps \u002F 16fps \u002F 19fps\u003C\u002Fspan>\u003C\u002Fdiv>\u003C\u002Fdiv>                     |\n| LLM           | ❌              | ❌           | \u003Cspan class=\"strong2\">Qwen\u002FDeepSeek 0.5B(fftf: 640ms, 9 tokens\u002Fs)\u003Cbr>Qwen\u002FDeepSeek 1.5B(fftf: 1610ms, 4 tokens\u002Fs) \u003Cbr>VLM(InterVL 1B) \u003Cbr>Mode models\u003C\u002Fspan> |\n| OpenMV算法 | \u003Cdiv class=\"comment\">测试图片参考\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixPy\u002Ftree\u002Fmain\u002Fprojects\u002Fapp_benchmark\">Benchmark APP\u003C\u002Fa>  |                    \u003Cdiv class=\"comment\">测试图片参考\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixPy\u002Ftree\u002Fmain\u002Fprojects\u002Fapp_benchmark\">Benchmark APP\u003C\u002Fa>\u003Cbr>测试日期：2025.8.22，更新可能有优化\u003C\u002Fdiv>              |       \u003Cdiv class=\"comment\">测试图片参考\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixPy\u002Ftree\u002Fmain\u002Fprojects\u002Fapp_benchmark\">Benchmark APP\u003C\u002Fa>\u003Cbr>测试日期：2025.8.22，更新可能有优化\u003C\u002Fdiv>                             |\n|   \u003Cdiv class=\"right second\">二值化\u003C\u002Fdiv>  | 灰度320x240: 7.4ms (135fps)\u003Cbr>灰度640x480: ❌\u003Cbr>RGB 320x240: 11.3ms (88.5fps)\u003Cbr>RGB 640x480: ❌ | 灰度320x240: **3.1ms (326fps)**\u003Cbr>灰度640x480: **11ms (90fps)**\u003Cbr>RGB 320x240: **13.2ms (75fps)**\u003Cbr>RGB 640x480: **52.8ms (18fps)**        | 灰度320x240: \u003Cspan class=\"strong2\">1.3ms (799fps)\u003C\u002Fspan> \u003Cbr>灰度640x480: \u003Cspan class=\"strong2\">4.8ms (206fps)\u003C\u002Fspan>\u003Cbr>RGB 320x240: \u003Cspan class=\"strong2\">3.4ms (294fps)\u003C\u002Fspan>\u003Cbr>RGB 640x480: \u003Cspan class=\"strong2\">13.3ms (75fps)\u003C\u002Fspan> |\n|   \u003Cdiv class=\"right second\">寻找斑点\u003C\u002Fdiv>        | 320x240: 8.8ms (114fps) \u003Cbr>640x480: ❌| 320x240: **7ms (143fps)**  \u003Cbr>640x480: **20ms (50fps)**         | 320x240: \u003Cspan class=\"strong2\">3.7ms (271fps)\u003C\u002Fspan>\u003Cbr>640x480: \u003Cspan class=\"strong2\">11.1ms (89fps)\u003C\u002Fspan>  |\n|   \u003Cdiv class=\"right second\">单通道直方图\u003C\u002Fdiv>  | 320x240: **7.7ms (130fps)**\u003Cbr>640x480: ❌ | 320x240: **10.9ms (91fps)**\u003Cbr>640x480: **42.8ms (23fps)**       | 320x240: \u003Cspan class=\"strong2\">1.5ms (661fps)\u003C\u002Fspan>\u003Cbr>640x480: \u003Cspan class=\"strong2\">5.9ms (168fps)\u003C\u002Fspan>    |\n|   \u003Cdiv class=\"right second\">QR码\u003C\u002Fdiv>        | 320x240: **130.8ms (7.6fps)** \u003Cbr>640x480: ❌| 640x480: 136.9ms (7fps)\u003Cbr>NPU加速：\u003Cbr>  320x240: **22.1ms (45fps)**\u003Cbr>  640x480: 57.6ms (17fps)  | 640x480: 57.9ms (17fps)\u003Cbr>NPU加速：\u003Cbr>  320x240: \u003Cspan class=\"strong2\">9.2ms (109fps)\u003C\u002Fspan>   \u003Cbr>  640x480: \u003Cspan class=\"strong2\">23.2ms (43fps)\u003C\u002Fspan> |\n| OpenCV算法     |   | \u003Cdiv class=\"comment\">测试图片参考\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixPy\u002Ftree\u002Fmain\u002Fprojects\u002Fapp_benchmark\">Benchmark APP\u003C\u002Fa>\u003Cbr>测试日期：2025.8.22，更新可能有优化\u003C\u002Fdiv>    | \u003Cdiv class=\"comment\">测试图片参考\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixPy\u002Ftree\u002Fmain\u002Fprojects\u002Fapp_benchmark\">Benchmark APP\u003C\u002Fa>\u003Cbr>测试日期：2025.8.22，更新可能有优化\u003C\u002Fdiv>  |\n|   \u003Cdiv class=\"right second\">二值化\u003C\u002Fdiv>             | ❌  | 灰度320x240: **2.2ms (463fps)**     \u003Cbr>灰度640x480: **7.1ms (140fps)** | 灰度320x240: \u003Cspan class=\"strong2\">0.1ms (8174fps)\u003C\u002Fspan>  \u003Cbr>灰度640x480: \u003Cspan class=\"strong2\">0.3ms (2959fps)\u003C\u002Fspan>  |\n|   \u003Cdiv class=\"right second\">灰度自适应二值化\u003C\u002Fdiv> | ❌  | 320x240: **5.8ms (171fps)**     \u003Cbr>640x480: **21.3ms (46fps)**  | 320x240: \u003Cspan class=\"strong2\">1.6ms (608fps)\u003C\u002Fspan>  \u003Cbr>640x480: \u003Cspan class=\"strong2\">6.3ms (159fps)\u003C\u002Fspan> |\n|   \u003Cdiv class=\"right second\">单通道直方图\u003C\u002Fdiv>       | ❌  | 320x240: **1ms (1000fps)**     \u003Cbr>640x480: **6.2ms (160fps)**   | 320x240: \u003Cspan class=\"strong2\">0.4ms (2308fps)\u003C\u002Fspan>  \u003Cbr>640x480: \u003Cspan class=\"strong2\">1.7ms (604fps)\u003C\u002Fspan>  |\n|   \u003Cdiv class=\"right second\">寻找轮廓\u003C\u002Fdiv>           | ❌  | 320x240: **2.8ms (351fps)**    \u003Cbr>640x480: **8.6ms (116fps)**   | 320x240: \u003Cspan class=\"strong2\">0.4ms (2286fps)\u003C\u002Fspan>  \u003Cbr>640x480: \u003Cspan class=\"strong2\">1.4ms (692fps)\u003C\u002Fspan>  |\n\n## Maix 生态系统\n\nMaixPy 不仅是一个 Python SDK，更拥有完整的生态系统，涵盖硬件、软件、工具、文档，甚至云平台等。\n请参阅下图：\n\n![](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fsipeed_MaixPy_readme_df583d1f54a4.png)\n\n\n## 谁在使用 MaixPy？\n\n* **AI 算法工程师**，希望将 AI 模型部署到嵌入式设备上。\n> MaixPy 提供简单易用的 API 来访问 NPU，并配有文档帮助您开发 AI 模型。\n* **STEM 教师**，希望向学生教授 AI 和嵌入式知识。\n> MaixPy 提供易于使用的 API、PC 工具以及在线 AI 训练服务……让您专注于教授 AI，而无需纠结于硬件和复杂的软件操作。\n* **创客**，想制作一些酷炫的项目，但又不想学习复杂的硬件和软件。\n> MaixPy 提供 Python API，您只需掌握基本的 Python 语法即可。MaixPy 的 API 非常简单易用，几分钟内就能完成一个项目。\n* **工程师**，希望快速实现项目原型。\n> MaixPy 构建项目非常便捷，并提供相应的 C++ SDK，因此您可以直接使用 MaixPy 将 Python 代码部署或转换为 C++，只需几分钟即可完成。\n* **学生**，希望学习 AI 和嵌入式开发。\n> 我们提供了丰富的文档和教程，以及大量的开源代码，帮助您找到合适的学习路径，逐步成长。从简单的 Python 编程到 `视觉`、`AI`、`音频`、`Linux`、`RTOS` 等领域。\n* **企业**，希望开发 AI 视觉产品，但缺乏时间和专业工程师来构建复杂的嵌入式系统。\n> 您可以使用 MaixPy，甚至通过图形化编程来开发产品，无需额外的人力和时间。例如，根据需求为生产线添加 AI 质量检测系统，或为办公室配备 AI 安全监控设备。\n* **竞赛选手**，希望在比赛中取得好成绩。\n> MaixPy 集成了多种功能且易于使用，能够加快您的工作进度，在有限的时间内助您赢得比赛。已有许多参赛者凭借 MaixPy 在各类竞赛中脱颖而出。\n\n\n## 编译源代码\n\n如果您希望从源代码编译 MaixPy 固件，请参考 [编译 MaixPy 源代码](https:\u002F\u002Fwiki.sipeed.com\u002Fmaixpy\u002Fdoc\u002Fen\u002Fsource_code\u002Fbuild.html) 页面。\n\n\n## 许可证\n\n本仓库中的所有文件均遵循 [Apache License 2.0 Sipeed Ltd.](.\u002FLICENSE) 协议，第三方库除外，这些库可能有各自的许可证。\n\n\n## 社区\n\n* 项目分享：[maixhub.com\u002Fshare](https:\u002F\u002Fmaixhub.com\u002Fshare)\n* 应用分享：[maixhub.com\u002Fapp](https:\u002F\u002Fmaixhub.com\u002Fapp)\n* 讨论区：[maixhub.com\u002Fdiscussion](https:\u002F\u002Fmaixhub.com\u002Fdiscussion)\n* QQ 群：862340358\n* Telegram：[t.me\u002Fmaixpy](https:\u002F\u002Ft.me\u002Fmaixpy)\n* Github 问题：[github.com\u002Fsipeed\u002Fmaixpy\u002Fissues](https:\u002F\u002Fgithub.com\u002Fsipeed\u002Fmaixpy\u002Fissues)","# MaixPy v4 快速上手指南\n\nMaixPy (v4) 是一款专为边缘计算设备设计的 Python 开发框架，结合强大的硬件（如 MaixCAM 系列），可轻松实现 AI 视觉、音频处理及大模型应用。\n\n## 1. 环境准备\n\n### 硬件要求\n推荐使用以下官方硬件平台以获得最佳体验：\n*   **MaixCAM2**: 高性能首选（1.2GHz A53, 3.2Tops NPU, 支持运行 LLM\u002FVLM）。\n*   **MaixCAM \u002F MaixCAM-Pro**: 高性价比选择（1GHz RISC-V, 1Tops NPU, 运行 Linux\u002FRTOS 双系统）。\n*   *注：旧的 K210 和 V831 芯片因内存和性能限制，建议升级至 MaixCAM 系列。*\n\n### 软件要求\n*   **操作系统**: Windows 10\u002F11, macOS, 或 Linux (Ubuntu 推荐)。\n*   **开发工具**: \n    *   **MaixVision IDE** (推荐): 官方集成开发环境，提供一键烧录、调试和模型部署功能。\n    *   或任意支持串口通信的文本编辑器 + 终端。\n*   **网络**: 用于下载模型和连接 MaixHub 在线训练平台。\n\n## 2. 安装步骤\n\n### 方式一：使用 MaixVision IDE（推荐新手）\n1.  访问 [MaixVision 下载页面](https:\u002F\u002Fwiki.sipeed.com\u002Fen\u002Fmaixvision) 下载并安装最新版软件。\n2.  通过 USB 线将 MaixCAM 设备连接至电脑。\n3.  打开 MaixVision，软件会自动识别设备。\n4.  在 IDE 中直接编写代码或打开示例项目，点击“运行”即可将脚本下发至设备执行。\n\n### 方式二：通过 PyPI 安装核心库（适用于高级用户\u002F自定义构建）\n如果你需要在本地构建固件或使用纯 Python 包管理：\n\n```bash\npip install maixpy\n```\n\n*国内用户若下载缓慢，可使用清华源加速：*\n```bash\npip install maixpy -i https:\u002F\u002Fpypi.tuna.tsinghua.edu.cn\u002Fsimple\n```\n\n> **注意**：对于嵌入式设备端的完整系统镜像（包含 OS、驱动和运行时），通常直接使用官方提供的固件烧录工具（集成在 MaixVision 中）进行刷写，而非单纯通过 pip 安装。\n\n## 3. 基本使用\n\nMaixPy 的核心优势在于极简的 API 设计。以下是一个经典的 **AI 图像分类** 示例，仅需约 10 行代码即可实现摄像头采集、AI 推理和屏幕显示。\n\n### 示例：10 行代码实现 AI 物体识别\n\n确保设备已烧录包含 `mobilenetv2.mud` 模型的固件，或将模型文件上传至设备 `\u002Froot\u002Fmodels\u002F` 目录。\n\n```python\nfrom maix import camera, display, image, nn\n\n# 1. 加载预训练模型\nclassifier = nn.Classifier(model=\"\u002Froot\u002Fmodels\u002Fmobilenetv2.mud\")\n\n# 2. 初始化摄像头，分辨率自动匹配模型输入\ncam = camera.Camera(classifier.input_width(), classifier.input_height(), classifier.input_format())\n\n# 3. 初始化显示屏\ndisp = display.Display()\n\nwhile 1:\n    # 4. 读取图像\n    img = cam.read()\n    \n    # 5. 执行 AI 推理\n    res = classifier.classify(img)\n    max_idx, max_prob = res[0]\n    \n    # 6. 绘制结果并显示\n    msg = f\"{max_prob:5.2f}: {classifier.labels[max_idx]}\"\n    img.draw_string(10, 10, msg, image.COLOR_RED)\n    disp.show(img)\n```\n\n### 外设控制示例：串口通信\n\nMaixPy 同样简化了底层硬件操作，例如使用 UART 串口：\n\n```python\nfrom maix import uart\n\n# 列出可用串口设备\ndevices = uart.list_devices()\n\n# 初始化第一个串口，波特率 115200\nserial = uart.UART(devices[0], 115200)\n\n# 发送数据\nserial.write_str(\"hello world\")\n\n# 接收数据（超时 2000ms）\nprint(\"received:\", serial.read(timeout=2000))\n```\n\n### 进阶资源\n*   **在线模型训练**: 访问 [MaixHub](https:\u002F\u002Fmaixhub.com) 无需本地显卡即可一键训练并部署 AI 模型。\n*   **完整文档**: 查阅 [官方 Wiki](https:\u002F\u002Fwiki.sipeed.com\u002Fmaixpy\u002Fdoc\u002Fen\u002Findex.html) 获取 API 详解和更多硬件教程。","某初创团队需要在低成本嵌入式摄像头上快速部署一套实时垃圾分类识别系统，用于社区智能垃圾桶原型开发。\n\n### 没有 MaixPy 时\n- 开发人员必须精通 C\u002FC++ 底层驱动，编写数百行代码才能初始化摄像头和显示屏，硬件调试周期长达数周。\n- 移植 AI 模型需手动处理复杂的内存管理和数据格式转换，极易出现崩溃，且无法在设备端直接验证效果。\n- 缺乏统一的 Python 接口，每次调整识别逻辑都需要重新编译固件并烧录，迭代效率极低，严重拖慢产品上市节奏。\n- 外设（如串口通信）对接繁琐，需要查阅大量芯片手册编写寄存器操作代码，增加了项目维护难度。\n\n### 使用 MaixPy 后\n- 利用 MaixPy 简洁的 Python API，仅需约 10 行代码即可调用摄像头、加载模型并显示结果，将原型开发时间从数周缩短至半天。\n- 内置高效的 AI 推理引擎自动处理图像预处理与模型量化，开发者可直接在设备上运行 MobileNet 等模型，无需关心底层内存细节。\n- 支持热更新脚本，修改识别逻辑或阈值后无需重新编译固件，即时运行查看效果，极大提升了算法调优的敏捷性。\n- 通过统一的 `uart` 等模块接口，轻松实现与主控板的串口通信，几行代码即可完成数据上报，降低了系统集成门槛。\n\nMaixPy 让开发者能用熟悉的 Python 语言在边缘设备上像搭积木一样快速构建高性能 AI 应用，彻底打破了嵌入式开发的技能壁垒。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fsipeed_MaixPy_a71a3139.png","sipeed","Sipeed","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002Fsipeed_ece27dc4.png","AIoT opensource hardware platform",null,"support@sipeed.com","SipeedIO","sipeed.com","https:\u002F\u002Fgithub.com\u002Fsipeed",[82,86,90,94,98,102,106,110,114,118],{"name":83,"color":84,"percentage":85},"Python","#3572A5",88.8,{"name":87,"color":88,"percentage":89},"Shell","#89e051",4.5,{"name":91,"color":92,"percentage":93},"CMake","#DA3434",2.2,{"name":95,"color":96,"percentage":97},"C++","#f34b7d",1.7,{"name":99,"color":100,"percentage":101},"Jupyter Notebook","#DA5B0B",1.1,{"name":103,"color":104,"percentage":105},"JavaScript","#f1e05a",1,{"name":107,"color":108,"percentage":109},"CSS","#663399",0.5,{"name":111,"color":112,"percentage":113},"HTML","#e34c26",0.2,{"name":115,"color":116,"percentage":117},"Makefile","#427819",0,{"name":119,"color":120,"percentage":117},"C","#555555",723,106,"2026-04-14T08:03:17","NOASSERTION",4,"Linux (BuildRoot\u002FUbuntu), RTOS","无需独立 GPU，板载 NPU (MaixCAM: 1 Tops@INT8, MaixCAM2: 3.2 Tops@INT8)","最低 256MB (MaixCAM), 推荐 1GB\u002F4GB (MaixCAM2)",{"notes":130,"python":131,"dependencies":132},"该工具主要运行在嵌入式硬件平台（MaixCAM\u002FMaixCAM2）上，而非通用 PC 操作系统。开发需配合专用硬件，支持 Linux 和 RTOS 双系统架构。提供 MaixVision 桌面端 IDE 辅助开发，AI 模型训练可通过 MaixHub 在线平台完成。旧版 K210\u002FV831 芯片因性能限制建议升级至新平台。","Python 3",[133,134,135,136],"maixpy","opencv","numpy","scipy (MaixCAM2)",[15,14,13],[139,140,133,141,142,143,144,145,146],"ai","edge-ai","ai-vision","embedded","monitor","maixcam","maixcam2","maixcam-pro","2026-03-27T02:49:30.150509","2026-04-15T04:16:42.495350",[150,155,160,165,170,175,180],{"id":151,"question_zh":152,"answer_zh":153,"source_url":154},33640,"MaixCAM 连接开放式（无密码）WiFi 失败怎么办？","确保不要删除 '\u002Fboot\u002Fwifi.sta' 文件。如果该文件缺失，初始化脚本中的判断逻辑会跳过配置步骤，导致无法连接。重新创建该文件即可解决。此外，官方后续会优化脚本，当密码为空时自动添加 'key_mgmt=NONE' 配置。","https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixPy\u002Fissues\u002F45",{"id":156,"question_zh":157,"answer_zh":158,"source_url":159},33641,"MaixCAM 播放 WAV 音频结束后有电流声或杂音如何解决？","这是内核 DMA 内存池的 Bug 导致的。请在 MaixCAM-Pro 上刷入 v4.10.3 或更高版本的固件镜像，该问题已在该版本中修复。","https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixPy\u002Fissues\u002F73",{"id":161,"question_zh":162,"answer_zh":163,"source_url":164},33642,"自定义模型在 MaixCAM 上推理速度极慢（相比 YOLO）怎么办？","尝试更改模型导出时的输出节点（output node），选择更靠前的层作为输出可能显著提升速度。同时，尝试导出为 BF16 或 INT8 格式。注意：如果在 MaixPy 中使用，不要在 model_deploy.py 命令中添加 '--quant_input' 参数。","https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixPy\u002Fissues\u002F107",{"id":166,"question_zh":167,"answer_zh":168,"source_url":169},33643,"遇到 'ModuleNotFoundError: No module named Maix' 错误，文档代码无法运行？","这是因为文档对应的是旧版 MaixPy-V1（适用于 K210 等芯片），而 MaixCAM 使用的是新版 MaixPy (V4)。MaixCAM 不支持 MaixPy-V1 的模块导入方式。如果使用视觉功能请使用 MaixCAM 及其对应的 API；若必须使用旧版音频等特定模块，可能需要换用 K210 设备。","https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixPy\u002Fissues\u002F44",{"id":171,"question_zh":172,"answer_zh":173,"source_url":174},33644,"实时拍照并推流到电脑时黑屏或无法传输视频流？","这通常是供电问题。请尝试直接使用电脑的 USB 接口为 MaixCAM 供电，因为电脑 USB 电压更稳定。使用其他电源适配器可能导致电压不稳定或不足，从而导致无法传输视频流。另外，如果提示 ffmpeg 版本过低，可以考虑升级 SDK 或手动编译新版本 ffmpeg。","https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixPy\u002Fissues\u002F144",{"id":176,"question_zh":177,"answer_zh":178,"source_url":179},33645,"编译系统镜像时出现 'fuse: mountpoint is not empty' 或网络下载错误？","如果是网络相关错误（如拉取资源失败），通常是因为国内网络环境限制。需要在终端开启代理（科学上网）才能正常下载位于 GitHub 上的资源。编译环境本身在 GitHub Actions 中已测试通过，本地编译请检查网络连接。","https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixPy\u002Fissues\u002F98",{"id":181,"question_zh":182,"answer_zh":183,"source_url":184},33646,"将 ONNX 模型转换为 MaixCAM 分类模型时，是否需要移除 Softmax 层？","是的，建议在转换前移除模型中的 Softmax 层。对于分类任务，应取 Softmax 层之前的输出作为模型输出。如果在转换时包含 Softmax 层，可能会影响推理结果的准确性或格式。具体转换指南可参考官方文档。","https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixPy\u002Fissues\u002F177",[186,191,196,201,206,211,216,221,226,231,236,240,245,250,255,259,264,269,274,279],{"id":187,"version":188,"summary_zh":189,"released_at":190},257928,"v4.6.0","## New features\r\n* Support [OCR(image text recognize)](https:\u002F\u002Fwiki.sipeed.com\u002Fmaixpy\u002Fdoc\u002Fzh\u002Fvision\u002Focr.html).\r\n* Support speech recognitioin.\r\n* Support new hardware MaixCAM-Pro.\r\n\r\nhttps:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002Fe00cc47e-f7df-4c8c-9bb6-4655d45fa081\r\n\r\n## Attention\r\n\r\n**Burn system image(`maixcam_*.img.xz`(for device MaixCAM) or `maixcam-pro_*.img.xz`(for device MaixCAM-Pro) file below) to use this version because it have camera driver update in system, and don't forget to install runtime in first bootup, old version system DO NOT upgrade runtime**\r\n**烧录系统镜像(下面的`maixcam_*.img.xz`(设备 MaixCAM 用)或`maixcam-pro_*.img.xz`(设备 MaixCAM-Pro 用)文件），以及第一次开机后记得安装运行库，才能正常使用此版本**\r\n\r\n## Downloads\r\n\r\n| github release assets (recommend!) |  Sourceforge  | Pypi.org |\r\n| --- | --- | --- |\r\n| ![GitHub release (by tag)](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fdownloads\u002Fsipeed\u002Fmaixpy\u002Fv4.6.0\u002Ftotal) Assets below    ↓↓↓↓↓↓↓↓↓  点击下方 Assets  |  [![SourceForge](https:\u002F\u002Fimg.shields.io\u002Fsourceforge\u002Fdt\u002Fmaixpy\u002Fv4.6.0)](https:\u002F\u002Fsourceforge.net\u002Fprojects\u002Fmaixpy\u002Ffiles\u002Fv4.6.0\u002F) |![PyPI - Downloads](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fdm\u002Fmaixpy?label=pypi)|\r\n> 国内下载速度太慢也可以到 QQ 群 862340358 群文件下载","2024-09-26T11:32:52",{"id":192,"version":193,"summary_zh":194,"released_at":195},257914,"v4.12.5","MaixCAM 发布说明\n\n关于 MaixCAM 的更多信息，请参阅：https:\u002F\u002Fwiki.sipeed.com\u002Fmaixcam\n\n## 新特性\n* [MaixCAM] 支持 yolo26n\n\n## Bug 修复\n* [MaixCAM] 修复了摄像头检测错误传感器的问题\n* [MaixCAM] 修复了使用 Maixvison 时无法释放 venc 资源的问题\n* [MaixCAM] 修复了 Maixcam UI 显示不完整的问题\n\n## 注意事项\n**烧录系统镜像时，请注意 assets 文件夹中的以下文件：**\n* `maixcam_*.img.xz`：适用于 MaixCAM 设备\n* `maixcam-pro_*.img.xz`：适用于 MaixCAM-Pro 设备\n请务必**不要烧录错镜像文件**，否则可能导致屏幕烧屏（通常放置一段时间后会缓慢恢复）。\n\n更多详细信息请参考[文档](https:\u002F\u002Fwiki.sipeed.com\u002Fmaixpy\u002Fdoc\u002Fzh\u002Fbasic\u002Fupgrade.html)。\n\n> 烧录系统镜像，下面 assets 中的文件：\n> * `maixcam_*.img.xz`: 设备 MaixCAM 用\n> * `maixcam-pro_*.img.xz`: 设备 MaixCAM-Pro 用\n> 注意**不要烧录错镜像文件**，烧录错会导致屏幕烧屏（一般放置一段时间会缓慢恢复）\n\n更多说明请看[文档](https:\u002F\u002Fwiki.sipeed.com\u002Fmaixpy\u002Fdoc\u002Fzh\u002Fbasic\u002Fupgrade.html)\n\n## 下载\n\n| GitHub 发布资产（推荐！） | SourceForge | PyPI.org |\n| --- | --- | --- |\n| ![GitHub 下载量（所有资产、所有版本）](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fdownloads\u002Fsipeed\u002Fmaixpy\u002Ftotal?label=all%20downloads)  \u003Cbr>![GitHub 发布（按标签）](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fdownloads\u002Fsipeed\u002Fmaixpy\u002Fv4.12.1\u002Ftotal) 下方为资产    ↓↓↓↓↓↓↓↓↓  点击下方 Assets  |  [![SourceForge](https:\u002F\u002Fimg.shields.io\u002Fsourceforge\u002Fdt\u002Fmaixpy-mirror\u002Fv4.12.1)](https:\u002F\u002Fsourceforge.net\u002Fprojects\u002Fmaixpy-mirror\u002Ffiles\u002Fv4.12.1) |![PyPI - 下载量](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fdm\u002Fmaixpy?label=pypi)|\n> 国内下载速度太慢也可以到 QQ 群 862340358 群文件下载\n","2026-01-21T12:35:03",{"id":197,"version":198,"summary_zh":199,"released_at":200},257915,"v4.12.4","MaixCAM、MaixCAM2 发布说明\n\n关于 MaixCAM 的信息，请参阅：https:\u002F\u002Fwiki.sipeed.com\u002Fmaixcam  \n关于 MaixCAM2 的信息，请参阅：https:\u002F\u002Fwiki.sipeed.com\u002Fmaixcam2  \n\n## 新特性\n* [MaixCAM2] os04d10&sc850sl 支持 60 帧率  \n* [MaixCAM2] 支持 snesevice 模型  \n* [MaixCAM2] 支持 somlvlm  \n* [MaixCAM2] 支持 qwen3vlm  \n\n## Bug 修复\n* 修复启动器死锁问题  \n* 修复 RTMP 音频编码器 bug  \n* [MaixCAM2] 修复无法解码带有 B 帧的 H.264 视频的问题  \n* [MaixCAM2] 为 maixcam2 添加 app_tof_camera 支持  \n\n## 应用更新\n* [MaixCAM2] 新增 thermal256 应用  \n\n## 注意事项\n**烧录系统镜像时，请使用 assets 文件夹中的以下文件：**  \n* `maixcam_*.img.xz`：适用于 MaixCAM 设备。  \n* `maixcam-pro_*.img.xz`：适用于 MaixCAM-Pro 设备。  \n* `maixcam2_*.img.7z.xxx`：适用于 MaixCAM2 设备。  \n* `maixcam2_*.axp`：MaixCAM2 的恢复镜像，烧录方法请参考 [文档](https:\u002F\u002Fwiki.sipeed.com\u002Fhardware\u002Fzh\u002Fmaixcam\u002Fmaixcam2_os.html)。  \n请注意 **不要烧录错误的镜像**，否则可能导致屏幕烧屏（通常放置一段时间后会缓慢恢复）。  \n\nMaixCAM2 镜像的下载方法请参见 [这里](https:\u002F\u002Fwiki.sipeed.com\u002Fhardware\u002Fen\u002Fmaixcam\u002Fmaixcam2_os.html)。  \n\n更多信息请参考 [文档](https:\u002F\u002Fwiki.sipeed.com\u002Fmaixpy\u002Fdoc\u002Fzh\u002Fbasic\u002Fupgrade.html)。  \n\n> 烧录系统镜像，下面 assets 中的文件：  \n> * `maixcam_*.img.xz`: 设备 MaixCAM 用  \n> * `maixcam-pro_*.img.xz`: 设备 MaixCAM-Pro 用  \n> 注意**不要烧录错镜像文件**，烧录错会导致屏幕烧屏（一般放置一段时间会缓慢恢复）  \n\n更多说明请看[文档](https:\u002F\u002Fwiki.sipeed.com\u002Fmaixpy\u002Fdoc\u002Fzh\u002Fbasic\u002Fupgrade.html)。  \n\n## 下载链接\n\n| GitHub 发布资产（推荐！） | SourceForge | PyPI.org |\n| --- | --- | --- |\n| ![GitHub 下载量（所有资产、所有版本）](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fdownloads\u002Fsipeed\u002Fmaixpy\u002Ftotal?label=all%20downloads)  \u003Cbr>![GitHub 发布（按标签）](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fdownloads\u002Fsipeed\u002Fmaixpy\u002Fv4.12.1\u002Ftotal) 下方 Assets    ↓↓↓↓↓↓↓↓↓  点击下方 Assets  |  [![SourceForge](https:\u002F\u002Fimg.shields.io\u002Fsourceforge\u002Fdt\u002Fmaixpy-mirror\u002Fv4.12.1)](https:\u002F\u002Fsourceforge.net\u002Fprojects\u002Fmaixpy-mirror\u002Ffiles\u002Fv4.12.1) |![PyPI - 下载量](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fdm\u002Fmaixpy?label=pypi)|\n> 国内下载速度太慢也可以到 QQ 群 862340358 群文件下载","2025-12-30T06:54:25",{"id":202,"version":203,"summary_zh":204,"released_at":205},257916,"v4.12.1","MaixCAM2 发布公告，有关 MaixCAM2 的信息，请参阅 https:\u002F\u002Fwiki.sipeed.com\u002Fmaixcam2\n\n\n## 新特性\n* [MaixCAM2] 全新平台 MaixCAM2，支持 LLM\u002FVLM\u002FASR\u002FTTS，并在多项算法中实现更快的运行速度。\n\n## Bug 修复\n\n* [MaixCAM] 修复自动启动应用卡死问题。\n\n## 应用更新\n\n* 新增基准测试 APP。\n* 新增 USB 手势控制 PC 鼠标 APP。\n* 新增 YOLO11 OBB APP。\n* [MaixCAM2] 新增 LLM 聊天 APP。\n* [MaixCAM2] 新增 VLM APP。\n* [MaixCAM2] 新增单目深度估计 APP。\n* [MaixCAM2] 新增 YOLO-World APP。\n\n\n## 注意事项\n**烧录系统镜像时，请使用 assets 文件夹中的以下文件：**\n* `maixcam_*.img.xz`：适用于 MaixCAM 设备。\n* `maixcam-pro_*.img.xz`：适用于 MaixCAM-Pro 设备。\n* `maixcam2_*.img.xz`：适用于 MaixCAM2 设备。\n* `maixcam2_*.axp`：MaixCAM2 的恢复镜像，请参考烧录 [文档](https:\u002F\u002Fwiki.sipeed.com\u002Fhardware\u002Fzh\u002Fmaixcam\u002Fmaixcam2_os.html)。\n请务必**不要烧录错误的镜像文件**，否则可能导致屏幕烧屏（通常放置一段时间后会缓慢恢复）。\n\n更多信息请参阅 [文档](https:\u002F\u002Fwiki.sipeed.com\u002Fmaixpy\u002Fdoc\u002Fzh\u002Fbasic\u002Fupgrade.html)。\n\n\n> 烧录系统镜像，下面 assets 中的文件：\n> * `maixcam_*.img.xz`: 设备 MaixCAM 用\n> * `maixcam-pro_*.img.xz`: 设备 MaixCAM-Pro 用\n> 注意**不要烧录错镜像文件**，烧录错会导致屏幕烧屏（一般放置一段时间会缓慢恢复）\n\n更多说明请看[文档](https:\u002F\u002Fwiki.sipeed.com\u002Fmaixpy\u002Fdoc\u002Fzh\u002Fbasic\u002Fupgrade.html)\n\n## 下载\n\n| GitHub 发布资产（推荐！） | SourceForge | PyPI.org |\n| --- | --- | --- |\n| ![GitHub 下载量（所有资产、所有版本）](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fdownloads\u002Fsipeed\u002Fmaixpy\u002Ftotal?label=all%20downloads)  \u003Cbr>![GitHub 发布版本（按标签）](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fdownloads\u002Fsipeed\u002Fmaixpy\u002Fv4.12.1\u002Ftotal) 下方为资产    ↓↓↓↓↓↓↓↓↓  点击下方 Assets  |  [![SourceForge](https:\u002F\u002Fimg.shields.io\u002Fsourceforge\u002Fdt\u002Fmaixpy\u002Fv4.12.1)](https:\u002F\u002Fsourceforge.net\u002Fprojects\u002Fmaixpy\u002Ffiles\u002Fv4.12.1\u002F) |![PyPI - 下载量](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fdm\u002Fmaixpy?label=pypi)|\n> 国内下载速度太慢也可以到 QQ 群 862340358 群文件下载","2025-09-30T11:48:03",{"id":207,"version":208,"summary_zh":209,"released_at":210},257917,"v4.11.9","测试版，遇到问题请到 issue 反馈，或者直接使用稳定版本固件\n\n## 新特性\n\n* 在 USB 升级模式和正常启动时，启动 LOGO 将有所不同。\n* **`image.Image()` 支持 `bg` 参数，用于设置背景或清空内存（默认情况下不清空内存以加快初始化速度。如果遇到 [随机背景内容](https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixPy\u002Fissues\u002F121) 的问题，请设置 `bg` 参数）。**\n* **[新增 IMU AHRS 支持](https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixCDK\u002Fcommit\u002F40bcfa4ac913fa63fd1fc950e0ac8b3cce5c4115)，请参考 [APP](https:\u002F\u002Fmaixhub.com\u002Fapp\u002F128) 和 [文档](https:\u002F\u002Fwiki.sipeed.com\u002Fmaixpy\u002Fdoc\u002Fzh\u002Fmodules\u002Fahrs.html)。**\n* **Maix 协议支持禁用，并且 [增加了 Maix 通信协议的 UART 端口选择功能](https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixCDK\u002Fcommit\u002F50be2300d184c52e4b0ec5deabe49aa1d3e2ab46)，现在您可以在 `settings` APP 中选择默认的 Maix 协议 UART。**\n* **相机支持将 `buff_num=1`，以获得更低的延迟图像，请参考 [文档](https:\u002F\u002Fwiki.sipeed.com\u002Fmaixpy\u002Fdoc\u002Fzh\u002Fvision\u002Fcamera.html)。**\n* **相机支持禁用 AWB 并手动设置 WB 增益**，请参考 [文档](https:\u002F\u002Fwiki.sipeed.com\u002Fmaixpy\u002Fdoc\u002Fzh\u002Fvision\u002Fcamera.html)。\n* [fs 添加 File.sync API](https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixCDK\u002Fcommit\u002Fa968572c445c525e5cf410f60777236e3fbc19e7)\n* [新增 USB 音频驱动支持](https:\u002F\u002Fgithub.com\u002Fsipeed\u002FLicheeRV-Nano-Build\u002Fcommit\u002F017c2724d2102c8af652019fc95667a868a25d76)。\n\n## 优化\n\n* **优化了 MaixCAM 开机时的 APP 启动速度，提速 2~5 秒。**\n* **工具链 `memcpy` 支持 RVV 加速，因此许多复制操作可能会更快，例如 `img.copy()` 等。**\n* [新的系统设置配置文件位于 `\u002Fboot\u002Fconfigs`。](https:\u002F\u002Fwiki.sipeed.com\u002Fmaixpy\u002Fdoc\u002Fzh\u002Fbasic\u002Fapp.html)\n* [优化了一些基础 API，如 log、memory_info、fs.join](https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixCDK\u002Fcommit\u002F3367869092c0b0542db8e7f0a320203b3b14a8eb)\n* [将 adc 引脚参数改为 id](https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixCDK\u002Fcommit\u002Fd117feebc487b216189c3a5a95469db62e878982)\n* [修复语音相关问题并优化启动速度](https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixCDK\u002Fcommit\u002F958d81ada7c4e3c7a7a0e46aa8be275e7e9a1279)\n\n## BUG 修复\n\n* OS04A10 90fps 模式消耗过多 CPU 资源，由于帧率低于 90fps，现已降至 80fps。\n* [修复 OS04A10 导致系统死机的问题](https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixCDK\u002Fcommit\u002F735ffa24ec1783ca07e542676e257d83812697ca)\n* [修复 get_pixel RGB888 获取错误整数数据的 bug](https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixCDK\u002Fcommit\u002Ffac17461e6e33c709b003703064a1b617271a9fa)\n* [修复 image.load PNG 文件的 bug](https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixCDK\u002Fcommit\u002F982a16396bdad687cc741a1c9e46109c0e74abd7)\n* [修复 sys.memory_info 溢出的 bug](https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixCDK\u002Fcommit\u002F9f10ae60c40d6fb2b2544a3e2de22c350d9c379b)\n* [修复：修正 Modbus 读取函数中的参数顺序](https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixCDK\u002Fcommit\u002F66a581d8dbac318f0e99e7c068664f2284cadb18)\n* 修复了开机有几率启动失败卡死在开机界面的问题，以及 MaixCAM-Pro 开机后功率太大会重启的问题，还有可能无法充电的问题。\n* 相比 v4.11.6，修复了人脸关键点和手部关键点检测无法运行的问题。\n\n## APP 更新\n\n* [增加图片列表的底部边距，以避免…]","2025-08-04T13:02:33",{"id":212,"version":213,"summary_zh":214,"released_at":215},257918,"v4.10.3","## 新特性\n\n* [[thermal32] 在融合图像中添加对 os04a10 的支持](https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixCDK\u002Fcommit\u002F89dfafc2d687f94935a0f60791db42e589a6d591)\n\n## 优化\n\n* [优化 Maix 通信协议监听器，现在退出更快，从而加快 UART 初始化](https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixCDK\u002Fcommit\u002Fb119fe5f7ddb88f37dcb483166bd028be8fdaefa)\n* [优化人脸表情示例的准确性](https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixPy\u002Fcommit\u002Ff8ce8ff5850722e1f3d8b191c8e75cb0053d8f7e)\n* [优化 maixcam 模型转换工具文档](https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixPy\u002Fcommit\u002F1f13afc52a8114eb3f9366c131fd16516b2e905d)\n\n## Bug 修复\n\n* [[tof100] 修复运行时错误](https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixPy\u002Fcommit\u002F4cad9f218f3cfe73a30ceaa6064d00e4aa4e2903)\n* [修复姿态识别：range() 不能接受浮点值](https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixPy\u002Fcommit\u002F140a66c688e9e65cd6c17e3b46fef218b136433b)\n* [修复 b26\u002Fb27 无法用作 GPIO 的问题](https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixCDK\u002Fcommit\u002F2e78cf739472d242f59d371e1943b8c4218c08f9)\n* [修复 b26\u002Fb27 无法用作 PWM 的问题](https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixCDK\u002Fcommit\u002F63d8e6c60105c97476ff703fb7a645d04a02eec6)\n* [修复 is_report 参数错误的协议问题](https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixCDK\u002Fcommit\u002F8e88e489ad221da91dce989eb06c4b811b28a343)\n* [修复首次以 channel=2 启动 PCM 时出现的段错误。](https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixCDK\u002Fcommit\u002F9257ba1b52c84382e7977a1d2f5a3f5e9668bb22)\n* 修复音频录制和播放相关问题：\n  * [修复播放器问题](https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixCDK\u002Fcommit\u002Fa5e3678ec7b5c2ea32058bc0ee0f0b7c50dc1230)\n  * [修复编码器记录时间不准确的问题](https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixCDK\u002Fcommit\u002F6f3b46dd0c90c27d86574ca4e5c7ab3812b8d990)\n  * [修复录制问题](https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixCDK\u002Fcommit\u002F31b6d482daae04ee903af29563da6bb7067509bd)\n\n## 注意事项\n**烧录系统镜像时，请注意 assets 文件夹中的以下文件：**\n* `maixcam_*.img.xz`：适用于 MaixCAM 设备。\n* `maixcam-pro_*.img.xz`：适用于 MaixCAM-Pro 设备。\n请务必**不要烧录错误的镜像文件**，否则可能导致屏幕烧屏（通常放置一段时间后会缓慢恢复）。\n\n更多信息请参阅[文档](https:\u002F\u002Fwiki.sipeed.com\u002Fmaixpy\u002Fdoc\u002Fzh\u002Fbasic\u002Fupgrade.html)。\n\n> 烧录系统镜像，下面 assets 中的文件：\n> * `maixcam_*.img.xz`: 设备 MaixCAM 用\n> * `maixcam-pro_*.img.xz`: 设备 MaixCAM-Pro 用\n> 注意**不要烧录错镜像文件**，烧录错会导致屏幕烧屏（一般放置一段时间会缓慢恢复）\n\n更多说明请看[文档](https:\u002F\u002Fwiki.sipeed.com\u002Fmaixpy\u002Fdoc\u002Fzh\u002Fbasic\u002Fupgrade.html)\n\n## 下载\n\n| GitHub 发布资产（推荐！） | SourceForge | PyPI.org |\n| --- | --- | --- |\n| ![GitHub 下载量（所有资产、所有版本）](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fdownloads\u002Fsipeed\u002Fmaixpy\u002Ftotal?label=all%20downloads)  \u003Cbr>![GitHub 发布（按标签）](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fdownloads\u002Fsipeed\u002Fmaixpy\u002Fv4.10.3\u002Ftotal) 下方资产    ↓↓↓↓↓↓↓↓↓  点击下方 Assets  |  [![SourceForge](https:\u002F\u002Fimg.shields.io\u002Fsourceforge\u002Fdt\u002Fmaixpy\u002Fv4.10.3)](https:\u002F\u002Fsourceforge.net\u002Fprojects\u002Fmaixpy\u002Ffiles\u002Fv4.10.3\u002F) |![PyPI - 下载量](https:\u002F\u002Fimg.shields.io\u002Fpy","2025-03-19T08:31:36",{"id":217,"version":218,"summary_zh":219,"released_at":220},257919,"v4.10.2","## 优化\n* 优化 TOF 模块\n* [os04a10 支持 720p@90FPS](https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixCDK\u002Fcommit\u002F7a4526318a4cbeab56937ee0bcb70d55d60389c1)\n\n\n## Bug 修复\n\n* 协议中 is_report 属性错误的 bug。\n* 修复首次以 channel=2 启动 PCM 时出现的段错误。\n* 修复音频播放器的 bug。\n* [修复编码器记录时间不准确的问题](https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixCDK\u002Fcommit\u002F6f3b46dd0c90c27d86574ca4e5c7ab3812b8d990)\n\n## 注意事项\n**烧录系统镜像时，请注意 assets 文件夹中的以下文件：**\n* `maixcam_*.img.xz`：适用于 MaixCAM 设备。\n* `maixcam-pro_*.img.xz`：适用于 MaixCAM-Pro 设备。\n请务必**不要烧录错镜像文件**，否则可能导致屏幕烧屏（通常放置一段时间后会缓慢恢复）。\n\n更多信息请参考[文档](https:\u002F\u002Fwiki.sipeed.com\u002Fmaixpy\u002Fdoc\u002Fzh\u002Fbasic\u002Fupgrade.html)。\n\n\n> 烧录系统镜像，下面 assets 中的文件：\n> * `maixcam_*.img.xz`: 设备 MaixCAM 用\n> * `maixcam-pro_*.img.xz`: 设备 MaixCAM-Pro 用\n> 注意**不要烧录错镜像文件**，烧录错会导致屏幕烧屏（一般放置一段时间会缓慢恢复）\n\n更多说明请看[文档](https:\u002F\u002Fwiki.sipeed.com\u002Fmaixpy\u002Fdoc\u002Fzh\u002Fbasic\u002Fupgrade.html)\n\n## 下载\n\n| GitHub 发布资产（推荐！） | SourceForge | PyPI.org |\n| --- | --- | --- |\n| ![GitHub 下载量（所有资产、所有版本）](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fdownloads\u002Fsipeed\u002Fmaixpy\u002Ftotal?label=all%20downloads)  \u003Cbr>![GitHub 发布版本（按标签）](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fdownloads\u002Fsipeed\u002Fmaixpy\u002Fv4.10.2\u002Ftotal) 下方资产    ↓↓↓↓↓↓↓↓↓  点击下方 Assets  |  [![SourceForge](https:\u002F\u002Fimg.shields.io\u002Fsourceforge\u002Fdt\u002Fmaixpy\u002Fv4.10.2)](https:\u002F\u002Fsourceforge.net\u002Fprojects\u002Fmaixpy\u002Ffiles\u002Fv4.10.2\u002F) |![PyPI - 下载量](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fdm\u002Fmaixpy?label=pypi)|\n> 国内下载速度太慢也可以到 QQ 群 862340358 群文件下载","2025-02-12T09:28:43",{"id":222,"version":223,"summary_zh":224,"released_at":225},257920,"v4.10.1","## 新特性\n\n* 增加了支持深度信息的人脸478个关键点检测功能，可靠且快速，请参阅[文档](https:\u002F\u002Fwiki.sipeed.com\u002Fmaixpy\u002Fdoc\u002Fzh\u002Fvision\u002Fface_landmarks.html)([英文](https:\u002F\u002Fwiki.sipeed.com\u002Fmaixpy\u002Fdoc\u002Fen\u002Fvision\u002Fface_landmarks.html))。\n![image](https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002F6f6ebc05-5824-4ec8-b2af-4862f46abf07)\n视频：\n\nhttps:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002F99bb0986-ad77-4cd6-bd21-e6ff5409b05d\n\n* 增加了人脸表情识别支持，默认支持7种类型，您也可以根据[文档](https:\u002F\u002Fwiki.sipeed.com\u002Fmaixpy\u002Fdoc\u002Fzh\u002Fvision\u002Fface_emotion.html)([英文](https:\u002F\u002Fwiki.sipeed.com\u002Fmaixpy\u002Fdoc\u002Fen\u002Fvision\u002Fface_emotion.html))进行自定义训练。\n![image](https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002F26b5dda7-35dc-4c0a-a8f4-a16fdb8091ec) \n![image](https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002F151f4eb4-3760-4944-9dd4-c6ce4c359f2b)\n\n\n视频：\n\nhttps:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002Fe7e7f6f7-1907-4bc2-9871-799d0f247e6b\n\n> 视频素材来自 https:\u002F\u002Fgithub.com\u002Foarriaga\u002Fface_classification\n\n* 所有YOLO检测API现在都支持`sort`参数，因此您可以根据目标大小对结果进行排序，只保留前几个较大的目标。\n* 增加了[人体姿态分类器](https:\u002F\u002Fwiki.sipeed.com\u002Fmaixpy\u002Fdoc\u002Fzh\u002Fvision\u002Fbody_pose_classification.html)([英文](https:\u002F\u002Fwiki.sipeed.com\u002Fmaixpy\u002Fdoc\u002Fen\u002Fvision\u002Fbody_pose_classification.html))。\n\n![image](https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002F9ffcffcf-3fe0-4be1-8daa-91f0a86f458a)\n\n## 应用程序\n\n* 增加了人脸关键点检测应用程序。\n* 增加了人脸表情识别应用程序。\n* 增加了人体姿态分类应用程序。\n* 相机应用程序优化了LED灯IO控制，现在您可以在`\u002Fboot\u002Fboard`中手动设置灯光IO。\n\n## Bug修复\n\n* **修复了设置自动启动程序时卡死的bug。**\n* [修复了ModBus的bug](https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixCDK\u002Fcommit\u002Ffe7ff536adefa382db6063fc09d2e6668b4f435b)\n\n## 优化\n\n* 优化了`maixtool`命令，使其与MaixVision生成的`app.yaml`文件完全兼容。\n\n## 注意事项\n**烧录系统镜像时，请注意assets文件夹中的以下文件：**\n* `maixcam_*.img.xz`: 用于MaixCAM设备。\n* `maixcam-pro_*.img.xz`: 用于MaixCAM-Pro设备。\n请务必**不要烧录错误的镜像文件**，否则可能会导致屏幕烧屏（通常放置一段时间后会缓慢恢复）。\n\n更多信息请参阅[文档](https:\u002F\u002Fwiki.sipeed.com\u002Fmaixpy\u002Fdoc\u002Fzh\u002Fbasic\u002Fupgrade.html)。\n\n\n> 烧录系统镜像，下面 assets 中的文件：\n> * `maixcam_*.img.xz`: 设备 MaixCAM 用\n> * `maixcam-pro_*.img.xz`: 设备 MaixCAM-Pro 用\n> 注意**不要烧录错镜像文件**，烧录错会导致屏幕烧屏（一般放置一段时间会缓慢恢复）\n\n更多说明请看[文档](https:\u002F\u002Fwiki.sipeed.com\u002Fmaixpy\u002Fdoc\u002Fzh\u002Fbasic\u002Fupgrade.html)\n\n## 下载\n\n| GitHub发布资产（推荐！） | Sourceforge | PyPI.org |\n| --- | --- | --- |\n| ![GitHub下载量（所有资产、所有版本）](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fdownloads\u002Fsipeed\u002Fmaixpy\u002Ftotal?label=all%20downloads)  \u003Cbr>![GitHub发布（按标签）](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fdownloads\u002Fsipeed\u002Fmaixpy\u002Fv4.10.1\u002Ftotal) 下方为资产    ↓↓↓↓↓↓↓↓↓  点击下方Assets","2025-01-14T03:16:40",{"id":227,"version":228,"summary_zh":229,"released_at":230},257921,"v4.10.0","请使用 4.10.1。","2025-01-13T06:58:48",{"id":232,"version":233,"summary_zh":234,"released_at":235},257922,"v4.9.3","## 新特性\n\n* 增加**手势关键点检测**功能，[中文文档](https:\u002F\u002Fwiki.sipeed.com\u002Fmaixpy\u002Fdoc\u002Fzh\u002Fvision\u002Fhand_landmarks.html)，[英文文档](https:\u002F\u002Fwiki.sipeed.com\u002Fmaixpy\u002Fdoc\u002Fen\u002Fvision\u002Fhand_landmarks.html)  \n视频：\n\nhttps:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002F34aa3146-fef5-4127-bfbd-7bb7755bf4d6\n\n\n* 增加**YOLO11-OBB**支持，现可检测带有角度的物体，[文档](https:\u002F\u002Fwiki.sipeed.com\u002Fmaixpy\u002Fdoc\u002Fzh\u002Fvision\u002Fdetect_obb.html)（[英文文档](https:\u002F\u002Fwiki.sipeed.com\u002Fmaixpy\u002Fdoc\u002Fen\u002Fvision\u002Fdetect_obb.html)）\n\n![image](https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002Fc7561e05-2548-49f9-a564-e555d369b162)\n\n* 增加对MaixCAM作为UVC摄像头的支持，并推出APP，可将MaixCAM用作USB摄像头。\n* 增加对热成像相机MLX90640的支持，相关硬件即将上架。  \n![image](https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002F30566f18-e31a-42ec-a6c6-1e1f261abf09)\n* 增加对TOF相机的支持，相关硬件即将上架。  \n![image](https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002Feed5683e-561b-4477-b139-170fd1ef94d2)\n* 支持在设置APP中禁用默认的CommProtocol（也可通过`\u002Fmaixapp\u002Fsys_conf.ini`进行设置）。\n* `image.to_format()`支持将NV21格式转换为灰度图。\n* `nn.NN.forward_image`现支持`hwc`通道格式，即也支持TensorFlow模型。您可以将TensorFlow模型转换为ONNX格式，再进一步转换为MaixCAM支持的格式。\n\n\n## Bug修复\n\n* 修复摄像头中出现多张人脸时的人脸识别错误。\n* 解决无网络情况下从RTC同步时间过晚的问题 #90。\n* 修复RTSP内存泄漏问题。\n* 修复MUD文件解析标签包含空格的问题。\n* 修复MaixCAM上的密钥初始化错误。\n\n## 优化\n\n* 将`nn.NN`类的`dual_buff`参数默认设置为`false`，以避免程序错误。\n* 在引脚映射中新增对PIN24\u002FPWM2的支持。（MaixCAM未引出该引脚）\n\n## MaixVision更新 v1.2.1\n\n* 正式支持MacOS系统，提供x64和arm64版本，并已通过开发者签名。\n\n## 注意事项\n**烧录系统镜像时，请注意以下assets文件夹中的文件：**\n* `maixcam_*.img.xz`：适用于MaixCAM设备。\n* `maixcam-pro_*.img.xz`：适用于MaixCAM-Pro设备。\n请务必**不要烧录错镜像文件**，否则可能导致屏幕烧屏（通常放置一段时间后会缓慢恢复）。\n\n更多信息请参阅[文档](https:\u002F\u002Fwiki.sipeed.com\u002Fmaixpy\u002Fdoc\u002Fzh\u002Fbasic\u002Fupgrade.html)。\n\n> 烧录系统镜像，下面 assets 中的文件：\n> * `maixcam_*.img.xz`: 设备 MaixCAM 用\n> * `maixcam-pro_*.img.xz`: 设备 MaixCAM-Pro 用\n> 注意**不要烧录错镜像文件**，烧录错会导致屏幕烧屏（一般放置一段时间会缓慢恢复）\n\n更多说明请看[文档](https:\u002F\u002Fwiki.sipeed.com\u002Fmaixpy\u002Fdoc\u002Fzh\u002Fbasic\u002Fupgrade.html)\n\n## 下载\n\n| GitHub发布资产（推荐！） | Sourceforge | PyPI.org |\n| --- | --- | --- |\n| ![GitHub release (by tag)](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fdownloads\u002Fsipeed\u002Fmaixpy\u002Fv4.9.3\u002Ftotal) 下方Assets ↓↓↓↓↓↓↓↓↓ 点击下方 Assets  |  [![SourceForge](https:\u002F\u002Fimg.shields.io\u002Fsourceforge\u002Fdt\u002Fmaixpy\u002Fv4.9.3)](https:\u002F\u002Fsourceforge.net\u002Fprojects\u002Fmaixpy\u002Ffiles\u002Fv4.9.3\u002F) |![PyPI - Downloads](https:\u002F\u002Fimg.shiel","2025-01-02T08:25:01",{"id":237,"version":238,"summary_zh":76,"released_at":239},257923,"v4.9.2","2024-12-20T09:41:04",{"id":241,"version":242,"summary_zh":243,"released_at":244},257924,"v4.8.2","## Bug修复\n\n* 应用程序相册播放视频时出现错误。\n* 修正MaixCAM-Pro相机应用中照片上的引脚图像。\n* 修复第二次设置自动启动应用程序失败的问题。\n* 修复在网络不可用时时间同步脚本卡死的问题。\n* 修复zbar无法解码中文的问题。\n* 修复应用程序HTTP文件浏览器打开大文件时出错的问题。\n\n## 优化\n\n* 将Modbus默认端口由5020改为502。\n* 相机应用支持添加时间戳。\n\n## 注意事项\n\n**烧录系统镜像时，请使用assets文件夹中的以下文件：**\n* `maixcam_*.img.xz`：适用于MaixCAM设备。\n* `maixcam-pro_*.img.xz`：适用于MaixCAM-Pro设备。\n请务必**不要烧录错误的镜像文件**，否则可能导致屏幕烧屏（通常放置一段时间后会缓慢恢复）。\n\n> 烧录系统镜像，下面 assets 中的文件：\n> * `maixcam_*.img.xz`: 设备 MaixCAM 用\n> * `maixcam-pro_*.img.xz`: 设备 MaixCAM-Pro 用\n> 注意**不要烧录错镜像文件**，烧录错会导致屏幕烧屏（一般放置一段时间会缓慢恢复）\n\n## 下载\n\n| GitHub发布资产（推荐！） | SourceForge | PyPI.org |\n| --- | --- | --- |\n| ![GitHub release (by tag)](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fdownloads\u002Fsipeed\u002Fmaixpy\u002Fv4.8.2\u002Ftotal) 下方Assets    ↓↓↓↓↓↓↓↓↓  点击下方 Assets  |  [![SourceForge](https:\u002F\u002Fimg.shields.io\u002Fsourceforge\u002Fdt\u002Fmaixpy\u002Fv4.8.2)](https:\u002F\u002Fsourceforge.net\u002Fprojects\u002Fmaixpy\u002Ffiles\u002Fv4.8.2\u002F) |![PyPI - Downloads](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fdm\u002Fmaixpy?label=pypi)|\n> 国内下载速度太慢也可以到 QQ 群 862340358 群文件下载","2024-12-05T12:08:16",{"id":246,"version":247,"summary_zh":248,"released_at":249},257925,"v4.8.0","## New features\r\n* Add **Modbus** support (based on UART and TCP), both slave and master mode are supported.\r\n* Key support **powerkey** of MaixCAM-Pro, and long press in launcher will pop out power option.\r\n* Add `maix.image.flip` API.\r\n* Camera APP support change **video bitrate**, and **Photos APP support play video** now.\r\n* Video record(Camera APP), RTSP, RTMP, video play(Photos APP) all **video related APPs support audio record and play**.\r\n* **`display.Display` now support select device fb(framebuffer)**, so you can port any screen in system as fb driver and use in display like if you want to use two screen or a SPI OLED screen.\r\n* **Support 7 inch touchscreen**.\r\n* LVGL based APPs **support mouse input**.\r\n![image](https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002F4199e2d8-0fa5-43fd-a115-03af856cf5e4)\r\n* Support a **MIPI to HDMI module**, so now we can use HDMI as screen.\r\n* Support audio player volume setting.\r\n* Add device config file in `\u002Fboot\u002Fboard`, so you can customize camera and display vflip and hmirror, and max LCD backlight power etc, refer to `\u002Fboot\u002Fboards\u002FREADME.txt` on device.\r\n* Launcher support I18N.\r\n![image](https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002F2579801f-613b-4194-a664-31f8a8acbcee)\r\n* **Add log file for APP**, you can find it in `\u002Fmaixapp\u002Flast_run.log`, so you can find bugs from log if your APP dead, use `examples\u002Ftools\u002Fshow_last_run_log.py` to see log.\r\n> so be carefully, print too much log may conduct large log file.\r\n* New QRCode API `maix.image.QRCodeDetector`, use **YOLOv8 to find QRCode, faster than classic one**.\r\n![image](https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002Fe12215b1-bc82-4da1-99ef-8a73285cd9d5)\r\n\r\n\r\n## BUG Fix\r\n* Fix nn.F.softmax execute error\r\n* Fix RTMP RTSP resource realease bug.\r\n\r\n\r\n## Optimization\r\n\r\n* Speech recognition APP switch mode faster now.\r\n* Remove upgrade MaixPy item from Settings APP to avoid user wrong upgrade.\r\n* Optimize API documentation click TOC jump(same id issue).\r\n\r\n## MaixVision Update v1.2.0\r\n\r\n* Support install runtime in MaixVision, so even MaixCAM no internet, you can also install runtime in MaixVision\r\n![image](https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002F27b3aabc-e51b-4234-b4b8-3e57cfc53092)\r\n* Add SSH terminal\r\n![image](https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002Fca40fcf1-d2c2-426e-b0b7-412fe2e812bc)\r\n* Add save image button on the right-top of image preview zone.\r\n* Image preview zone support show ROI info(width height etc.)\r\n* Support flush files button to sync files with system.\r\n![image](https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002Ff0d82584-450d-49dc-ac3c-4f5d2b1c8f35)\r\n* All py files can use Python code auto-suggestions and auto-completion.\r\n![image](https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002F0fd1e628-aa09-47e0-8c79-07d9da8bb96c)\r\n* Example code alway there now.\r\n* Optimize editor scroll to preview opened file names.\r\n![image](https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002F932aaef8-903c-468b-ad84-0ba5b3b0b743)\r\n* Optimize close files experience.\r\n* Fix LAB histogram value error bug.\r\n* Fix auto save will change AI model(binary) file bug.\r\n* Fix render fail and GUI disapear bug after click image preview zone some times.\r\n* Optimize multiple line error hint when pack APP.\r\n* Windows support customize installation path now.\r\n\r\n## Attention\r\n**Burning the system image, the following files in the assets folder:**\r\n* `maixcam_*.img.xz`: For the MaixCAM device.\r\n* `maixcam-pro_*.img.xz`: For the MaixCAM-Pro device.\r\nBe careful **not to burn the wrong image**, as burning the wrong one may cause screen burn-in (it usually recovers slowly after being left for a period of time).\r\n\r\n> 烧录系统镜像，下面 assets 中的文件：\r\n> * `maixcam_*.img.xz`: 设备 MaixCAM 用\r\n> * `maixcam-pro_*.img.xz`: 设备 MaixCAM-Pro 用\r\n> 注意**不要烧录错镜像文件**，烧录错会导致屏幕烧屏（一般放置一段时间会缓慢恢复）\r\n\r\n## Downloads\r\n\r\n| github release assets (recommend!) |  Sourceforge  | Pypi.org |\r\n| --- | --- | --- |\r\n| ![GitHub release (by tag)](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fdownloads\u002Fsipeed\u002Fmaixpy\u002Fv4.8.0\u002Ftotal) Assets below    ↓↓↓↓↓↓↓↓↓  点击下方 Assets  |  [![SourceForge](https:\u002F\u002Fimg.shields.io\u002Fsourceforge\u002Fdt\u002Fmaixpy\u002Fv4.8.0)](https:\u002F\u002Fsourceforge.net\u002Fprojects\u002Fmaixpy\u002Ffiles\u002Fv4.8.0\u002F) |![PyPI - Downloads](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fdm\u002Fmaixpy?label=pypi)|\r\n> 国内下载速度太慢也可以到 QQ 群 862340358 群文件下载","2024-11-29T09:44:17",{"id":251,"version":252,"summary_zh":253,"released_at":254},257926,"v4.7.8","## New features\r\n* Add **YOLO11** support(detect, segmentation, keypoint detection)\r\n![image](https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002F5698948e-4e73-4939-b3ee-c8c6640cce75)\r\n* Support New board [MaixCAM-Pro](https:\u002F\u002Fwiki.sipeed.com\u002Fmaixcam-pro) , include its PMU, IMU, RTC.\r\n![image](https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002Fb0abb7bd-e3e2-4eb3-aa16-430d9c2e8d5a)\r\n* SC035GS(Global shutter camera) support 180fps mode.\r\n* OS04A10 now support 40s exposure, GC4653 max 1\u002F3 s.\r\n* Camera support read RAW data.\r\n* Camera APP add LED light switch, RAW, digital zoomin.\r\n* New launcher UI\r\n![image](https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002Fa72a2137-7386-468a-b2d3-0c06ca8d6ba2)\r\n\r\n## BUG Fix\r\n\r\n* Fix speech recognition APP core dump bug\r\n\r\n## Optimization\r\n\r\n* optimize NN preprocess, now faster(e.g. 320x320 input, 12ms to 2ms)\r\n\r\n## Attention\r\n\r\n**Burn system image(`maixcam_*.img.xz`(for device MaixCAM) or `maixcam-pro_*.img.xz`(for device MaixCAM-Pro) file below) to use this version because it have camera driver update in system, and don't forget to install runtime in first bootup**\r\n**烧录系统镜像(下面的`maixcam_*.img.xz`(设备 MaixCAM 用)或`maixcam-pro_*.img.xz`(设备 MaixCAM-Pro 用)文件），以及第一次开机后记得联网安装运行库，才能正常使用此版本**\r\n\r\n## Downloads\r\n\r\n| github release assets (recommend!) |  Sourceforge  | Pypi.org |\r\n| --- | --- | --- |\r\n| ![GitHub release (by tag)](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fdownloads\u002Fsipeed\u002Fmaixpy\u002Fv4.7.8\u002Ftotal) Assets below    ↓↓↓↓↓↓↓↓↓  点击下方 Assets  |  [![SourceForge](https:\u002F\u002Fimg.shields.io\u002Fsourceforge\u002Fdt\u002Fmaixpy\u002Fv4.7.8)](https:\u002F\u002Fsourceforge.net\u002Fprojects\u002Fmaixpy\u002Ffiles\u002Fv4.7.8\u002F) |![PyPI - Downloads](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fdm\u002Fmaixpy?label=pypi)|\r\n> 国内下载速度太慢也可以到 QQ 群 862340358 群文件下载","2024-10-31T11:28:05",{"id":256,"version":257,"summary_zh":76,"released_at":258},257927,"v4.7.7","2024-10-24T10:00:45",{"id":260,"version":261,"summary_zh":262,"released_at":263},257929,"v4.5.1","## New Features\r\n* Self learning object detect and tracking, see [doc](https:\u002F\u002Fwiki.sipeed.com\u002Fmaixpy\u002Fdoc\u002Fzh\u002Fvision\u002Fself_learn_detector.html)\r\n\r\nhttps:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002F1a7b5056-982c-4458-8b19-b313ac9de7ff\r\n\r\n* Object tracking and object count, see [doc](https:\u002F\u002Fwiki.sipeed.com\u002Fmaixpy\u002Fdoc\u002Fzh\u002Fvision\u002Fobject_track.html)\r\n\r\nhttps:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002F7e2142c0-b29c-466d-90eb-a4f1228f1b4e\r\n\r\n* Video(H.264) hardware decode support, now you can play mp4 video files.\r\n\r\nhttps:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002F2fcbef54-792c-4549-88d9-f05dbe053dfc\r\n\r\n* USB HID simulate support(MaixCAM act as mouse\u002Fkeyboard\u002Ftouchpad).\r\n* Support OS04A10 camera sensor.\r\n* Add `time.fps` API. refer to `MaixPy\u002Fexamples\u002Fbasic\u002Fdemo_time.py`\r\n* Add convertion API between `tensor.Tensor` and `numpy.ndarray`, refer to `MaixPy\u002Fexamples\u002Fvision\u002Fai_vision\u002Fnn_forward.py`\r\n* Add bm8563 & qmi8658 & tmc2209 doc & example\r\n* Add  [http file browser APP](https:\u002F\u002Fmaixhub.com\u002Fapp\u002F59).\r\n* Default listenning uart command, now you can send command though uart to switch APP like [protocol doc](https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixCDK\u002Fblob\u002Fmain\u002Fdocs\u002Fdoc\u002Fconvention\u002Fprotocol.md) said.\r\n\r\n## BUG Fix\r\n* UART read block issue.\r\n* YOLOv5\u002FYOLOv8 score always 0.5 issue(iou bug).\r\n* App_store APP exit when scan QR code and no network, and settings the same.\r\n\r\n## Attention\r\n\r\nFor this version, if you want to use HID function, you must burn system, or you can upgrade maixpy and then upgrade runtime, but we remain recommend you directly re-burn system.\r\n\r\n**Burn system image(`*.img.xz` file below) to use this version because it have camera driver update in system, and don't forget to install runtime in first bootup, old version system DO NOT upgrade runtime**\r\n**烧录系统镜像(下面的`*.img.xz`文件），以及第一次开机后记得安装运行库，才能正常使用此版本**\r\n\r\n## Downloads\r\n\r\n| github release assets (recommend!) |  Sourceforge  | Pypi.org |\r\n| --- | --- | --- |\r\n| ![GitHub release (by tag)](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fdownloads\u002Fsipeed\u002Fmaixpy\u002Fv4.5.1\u002Ftotal) Assets below    ↓↓↓↓↓↓↓↓↓  点击下方 Assets  |  [![SourceForge](https:\u002F\u002Fimg.shields.io\u002Fsourceforge\u002Fdt\u002Fmaixpy\u002Fv4.5.1)](https:\u002F\u002Fsourceforge.net\u002Fprojects\u002Fmaixpy\u002Ffiles\u002Fv4.5.1\u002F) |![PyPI - Downloads](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fdm\u002Fmaixpy?label=pypi)|\r\n> 国内下载速度太慢也可以到 QQ 群 862340358 群文件下载\r\n\r\n\r\n","2024-09-05T11:00:33",{"id":265,"version":266,"summary_zh":267,"released_at":268},257930,"v4.4.22","## New Features\r\n* Add WiFi AP mode API and demo(in `MaixPy\u002Fexamples\u002Ftools` directory)\r\n* Now USB driver (virtual network card) can be auto recognized by Windows and MacOS and Linux, no need to install driver manually anymore!\r\n* Add USB setting in settings APP, you can select device mode or host mode(can use external USB devices like  USB camera), and virual network card selection.\r\n* Add timezone setting API and setting item in settings APP, and you can also find demo in `MaixPy\u002Fexamples\u002Ftools` directory.\r\n\r\n## Optimization\r\n* [optimize app.switch_app ignore same app switch](https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixCDK\u002Fcommit\u002F3fd0701f0cfbbabea4a3ee0766a6c8a005965b4c)\r\n* Optimize runtime installation to ensure old system use old runtime.\r\n* Optimize MaixPy upgrade UI, now you can exit without stuck.\r\n* Optimize builtin find_blob APP blue color threshold to find more blue color.\r\n* Increase kernel memory to 128MB, so now user space memory is 128MB(the 4.4.19 is 160MB) for better camera driver stability.\r\n* Add depracate warning to v1 API and optimize v1 examples' test image path.\r\n\r\n## BUG Fix\r\n* NN memory leak issue when use dual_buff mode.\r\n* Fix UART0 Tx pin write data error sometimes.\r\n* [Fix morph bug](https:\u002F\u002Fgithub.com\u002Fsipeed\u002FMaixCDK\u002Fcommit\u002F1069419a75c7a616b205654efb50e72c944c81f6)\r\n* Fix camera reopen bug.\r\n* Fix desktop-monitor APP run failed error.\r\n\r\n\r\n## Attention\r\n\r\n**You must to burn system image(`*.img.xz` file below) to use this version because it have camera driver update in system, and don't forget to install runtime in first bootup, old version system DO NOT upgrade runtime**\r\n**你必须烧录系统镜像(下面的`*.img.xz`文件），以及第一次开机后记得安装运行库，才能正常使用此版本**\r\n\r\n## Downloads\r\n\r\n| github release assets (recommend!) |  Sourceforge  | Pypi.org |\r\n| --- | --- | --- |\r\n| ![GitHub release (by tag)](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fdownloads\u002Fsipeed\u002Fmaixpy\u002Fv4.4.22\u002Ftotal) Assets below    ↓↓↓↓↓↓↓↓↓  点击下方 Assets  |  [![SourceForge](https:\u002F\u002Fimg.shields.io\u002Fsourceforge\u002Fdt\u002Fmaixpy\u002Fv4.4.22)](https:\u002F\u002Fsourceforge.net\u002Fprojects\u002Fmaixpy\u002Ffiles\u002Fv4.4.22\u002F) |![PyPI - Downloads](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fdm\u002Fmaixpy?label=pypi)|\r\n\r\n\r\n\r\n","2024-08-16T05:31:14",{"id":270,"version":271,"summary_zh":272,"released_at":273},257931,"v4.4.19","## Bug Fix\r\n\r\n* Cancel camera default 80fps mode, cause some camera not support this freq, if you want to use 80fps mode, set it in `camera.Camera(320, 240, fps=80)`\r\n* Camera add_channel bug fix\r\n* UART read function param `len` bug fix\r\n* Limit app desc must one line when app_store installing app\r\n\r\n\r\n## Attention\r\n\r\n**You must to burn system image(`*.img.xz` file below) to use this version because it have camera driver update in system, and don't forget to install runtime in first bootup, old version system DO NOT upgrade runtime**\r\n**你必须烧录系统镜像(下面的`*.img.xz`文件），以及第一次开机后记得安装运行库，才能正常使用此版本，旧版镜像请不要更新 运行库**\r\n\r\n## Downloads\r\n\r\n| github release assets (recommend!) |  Sourceforge  | Pypi.org |\r\n| --- | --- | --- |\r\n| ![GitHub release (by tag)](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fdownloads\u002Fsipeed\u002Fmaixpy\u002Fv4.4.19\u002Ftotal) Assets below    ↓↓↓↓↓↓↓↓↓  点击下方 Assets  |  [![SourceForge](https:\u002F\u002Fimg.shields.io\u002Fsourceforge\u002Fdt\u002Fmaixpy\u002Fv4.4.19)](https:\u002F\u002Fsourceforge.net\u002Fprojects\u002Fmaixpy\u002Ffiles\u002Fv4.4.19\u002F) |![PyPI - Downloads](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fdm\u002Fmaixpy?label=pypi)|\r\n\r\n","2024-07-28T13:45:00",{"id":275,"version":276,"summary_zh":277,"released_at":278},257932,"v4.4.15","## New Features\r\n* Add MaixHub client APP, now you can collect dataset by this APP and directly upload to MaixHub, and deploy model to MaixCAM in one click.\r\n* Add `image.Image.set_windowing` function.\r\n* Support now global shutter camera sms_sc035gs (only support 60FPS mode now, later will support 180fps~240fps)\r\n* Support 5 inch screen, and 7inch touchscreen, buy [here](https:\u002F\u002Fitem.taobao.com\u002Fitem.htm?spm=a1z10.1-c-s.w21412213-24984936557.35.5a6752b1Fv7lsZ&id=610352268976&skuId=5688232998030)\r\n \r\n## Optimization\r\n* Camera(GC4653) now default support 80FPS mode, if you want to use 30 or 60 fps, set `fps=30` or `fps=60` when construct `camera.Camera` object.\r\n* Change `image.Image.awb_mode` to `image.Image.set_awb` function.\r\n\r\n## Bug Fix\r\n* Fix camera image buffer error.\r\n* Fix cv2image grayscale image error\r\n\r\n## Attention\r\n\r\n**You must to burn system image(`*.img.xz` file below) to use this version because it have camera driver update in system, and don't forget to install runtime in first bootup, old version system DO NOT upgrade runtime**\r\n**你必须烧录系统镜像(下面的`*.img.xz`文件），以及第一次开机后记得安装运行库，才能正常使用此版本，旧版镜像请不要更新 运行库**\r\n\r\n\r\n## Downloads\r\n\r\n| github release assets (recommend!) |  Sourceforge  | Pypi.org |\r\n| --- | --- | --- |\r\n| ![GitHub release (by tag)](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fdownloads\u002Fsipeed\u002Fmaixpy\u002Fv4.4.15\u002Ftotal) Assets below    ↓↓↓↓↓↓↓↓↓  点击下方 Assets  |  [![SourceForge](https:\u002F\u002Fimg.shields.io\u002Fsourceforge\u002Fdt\u002Fmaixpy\u002Fv4.4.15)](https:\u002F\u002Fsourceforge.net\u002Fprojects\u002Fmaixpy\u002Ffiles\u002Fv4.4.15\u002F) |![PyPI - Downloads](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fdm\u002Fmaixpy?label=pypi)|\r\n\r\n","2024-07-26T18:01:41",{"id":280,"version":281,"summary_zh":282,"released_at":283},257933,"v4.4.12","## New Features\r\n\r\n* i18n now support load from yaml dict files.\r\n* Camera APP support record video(mp4 format),  you can find video in `\u002Fmaixapp\u002Fshare\u002Fvideo` directory of device.\r\n\r\n## Optimization\r\n\r\n* Camera(GC4653) support 60FPS mode now, and now fefault 60FPS(only support when resolution \u003C= 720p).\r\n> 60FPS mode's color has some different with 30FPS mode,  if you want to use 30FPS mode, please set `fps=30` in `camera.Camera()`.\r\n* AI model(NPU) support dual_buff mode and set it to default, now runing faster twice than before.\r\n> dual buff mode enabled by default, if you want to disable it, set arg `dual_buff = False` when initilize model like `nn.YOLOv5(model=\"“, dual_buff=False)\r\n* APP bootup\u002Fswitch faster now.\r\n* YOLOv5 and YOLOv8 object coordinate now ensure all value >= 0.\r\n* `disp.set_backlight(0)` now will close backlight now.\r\n\r\n\r\n## BUG Fix\r\n\r\n* Setting install runtime success but always show not install bug.\r\n* Face recoginizer run fail bug.\r\n* image.cv2image and image.image2cv memory leak bug\r\n* RTSP optimize, now you can play it with `ffplay rtsp:\u002F\u002F10.228.104.1:8554\u002Flive` or `mpv rtsp:\u002F\u002F10.228.104.1:8554\u002Flive`\r\n\r\n\r\n## Downloads\r\n\r\n| github release assets (recommend!) |  Sourceforge  | Pypi.org |\r\n| --- | --- | --- |\r\n| ![GitHub release (by tag)](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fdownloads\u002Fsipeed\u002Fmaixpy\u002Fv4.4.12\u002Ftotal) Assets below    ↓↓↓↓↓↓↓↓↓  点击下方 Assets  |  [![SourceForge](https:\u002F\u002Fimg.shields.io\u002Fsourceforge\u002Fdt\u002Fmaixpy\u002Fv4.4.12)](https:\u002F\u002Fsourceforge.net\u002Fprojects\u002Fmaixpy\u002Ffiles\u002Fv4.4.12\u002F) |![PyPI - Downloads](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fdm\u002Fmaixpy?label=pypi)|\r\n\r\n\r\n","2024-07-18T06:20:35"]