[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"similar-margaretmz--awesome-tensorflow-lite":3,"tool-margaretmz--awesome-tensorflow-lite":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":77,"owner_email":76,"owner_twitter":72,"owner_website":78,"owner_url":79,"languages":76,"stars":80,"forks":81,"last_commit_at":82,"license":83,"difficulty_score":84,"env_os":85,"env_gpu":86,"env_ram":87,"env_deps":88,"category_tags":97,"github_topics":98,"view_count":32,"oss_zip_url":76,"oss_zip_packed_at":76,"status":17,"created_at":119,"updated_at":120,"faqs":121,"releases":122},7457,"margaretmz\u002Fawesome-tensorflow-lite","awesome-tensorflow-lite","An awesome list of TensorFlow Lite models, samples, tutorials, tools and learning resources.","awesome-tensorflow-lite 是一个专为 TensorFlow Lite 生态打造的精选资源清单，旨在帮助开发者轻松将人工智能模型部署到手机、嵌入式设备等边缘终端。随着全球超过 40 亿台设备运行该技术，如何在资源受限的环境下高效落地 AI 成为关键挑战。这份清单通过系统整理社区贡献的优质模型、示例代码、实用工具及学习教程，有效解决了开发者在寻找可靠参考实现和入门资料时面临的碎片化难题。\n\n无论是刚接触移动端 AI 的初学者，还是经验丰富的工程师与研究人员，都能从中获益。内容覆盖计算机视觉（如图像分类、目标检测）、自然语言处理、语音识别、推荐系统乃至游戏开发等多个领域。其独特亮点在于不仅提供了现成的预训练模型库，还收录了基于 MLIR 的新转换器、支持快速定制模型的 Model Maker、以及实现端侧个性化训练等前沿技术指南。此外，清单还包含了详细的插件 SDK 信息和多形式的学习资源（博客、书籍、视频），让用户能一站式获取从理论到实践的全方位支持，加速智能应用的开发与迭代。","\u003Cp align=\"center\">\n \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmargaretmz_awesome-tensorflow-lite_readme_d02815ab1b45.png\" alt=\"awesome tflite\" width=\"500\">\n\u003C\u002Fp>\n\n\u003C!-- omit in toc -->\n# Awesome TensorFlow Lite [![Awesome](https:\u002F\u002Fawesome.re\u002Fbadge.svg)](https:\u002F\u002Fawesome.re) [![PRs Welcome](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FPRs-welcome-brightgreen.svg?style=flat-square)](http:\u002F\u002Fmakeapullrequest.com) [![Twitter](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FTwitter-%40margaretmz-blue)](https:\u002F\u002Ftwitter.com\u002Fmargaretmz)\n\n[TensorFlow Lite](https:\u002F\u002Fwww.tensorflow.org\u002Flite) is a set of tools that help convert and optimize TensorFlow models to run on mobile and edge devices. It's currently running on more than 4 billion devices! With TensorFlow 2.x, you can train a model with tf.Keras, easily convert a model to .tflite and deploy it; or you can download a pretrained TensorFlow Lite model from the model zoo.\n\nThis is an awesome list of TensorFlow Lite models with sample apps, helpful tools and learning resources -\n* Showcase what the community has built with TensorFlow Lite\n* Put all the samples side-by-side for easy reference\n* Share knowledge and learning resources\n\nPlease submit a PR if you would like to contribute and follow the guidelines [here](CONTRIBUTING.md).\n\n\u003C!-- omit in toc -->\n ## Contents\n- [Past announcements:](#past-announcements)\n- [Models with samples](#models-with-samples)\n - [Computer vision](#computer-vision)\n - [Classification](#classification)\n - [Detection](#detection)\n - [Segmentation](#segmentation)\n - [Style Transfer](#style-transfer)\n - [Generative](#generative)\n - [Post estimation](#post-estimation)\n - [Other](#other)\n - [Text](#text)\n - [Speech](#speech)\n - [Recommendation](#recommendation)\n - [Game](#game)\n- [Model zoo](#model-zoo)\n - [TensorFlow Lite models](#tensorflow-lite-models)\n - [TensorFlow models](#tensorflow-models)\n- [Ideas and Inspiration](#ideas-and-inspiration)\n- [ML Kit examples](#ml-kit-examples)\n- [Plugins and SDKs](#plugins-and-sdks)\n- [Helpful links](#helpful-links)\n- [Learning resources](#learning-resources)\n - [Blog posts](#blog-posts)\n - [Books](#books)\n - [Videos](#videos)\n - [Podcasts](#podcasts)\n - [MOOCs](#moocs)\n\n## Past announcements:\nHere are some past feature annoucements of TensorFlow Lite:\n* [Announcement of the new converter](https:\u002F\u002Fgroups.google.com\u002Fa\u002Ftensorflow.org\u002Fd\u002Fmsg\u002Ftflite\u002FZ_h7706dt8Q\u002FsNrjPj4yGgAJ) - [MLIR](https:\u002F\u002Fmedium.com\u002Ftensorflow\u002Fmlir-a-new-intermediate-representation-and-compiler-framework-beba999ed18d)-based and enables conversion of new classes of models such as Mask R-CNN and Mobile BERT etc., supports functional control flow and better error handling during conversion. Enabled by default in the nightly builds\\.\n* [Android Support Library](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Ftflite-support\u002Ftree\u002Fmaster\u002Ftensorflow_lite_support\u002Fjava) - Makes mobile development easier ([Android](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fexamples\u002Fblob\u002Fmaster\u002Flite\u002Fexamples\u002Fimage_classification\u002Fandroid\u002FEXPLORE_THE_CODE.md) sample code).\n* [Model Maker](https:\u002F\u002Fwww.tensorflow.org\u002Flite\u002Fguide\u002Fmodel_maker) - Create your custom [image & text](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fexamples\u002Ftree\u002Fmaster\u002Ftensorflow_examples\u002Flite\u002Fmodel_maker) classification models easily in a few lines of code. See below the Icon Classifier for a tutorial by the community.\n* [On-device training](https:\u002F\u002Fblog.tensorflow.org\u002F2019\u002F12\u002Fexample-on-device-model-personalization.html) - It is finally here! Currently limited to transfer learning for image classification only but it's a great start. See the official [Android](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fexamples\u002Fblob\u002Fmaster\u002Flite\u002Fexamples\u002Fmodel_personalization\u002FREADME.md) sample code and another one from the community ([Blog](https:\u002F\u002Faqibsaeed.github.io\u002Fon-device-activity-recognition) | [Android](https:\u002F\u002Fgithub.com\u002Faqibsaeed\u002Fon-device-activity-recognition)).\n* [Hexagon delegate](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Ftensorflow\u002Fblob\u002Fmaster\u002Ftensorflow\u002Flite\u002Fg3doc\u002Fperformance\u002Fhexagon_delegate.md) - How to use the Hexagon Delegate to speed up model inference on mobile and edge devices. Also see blog post [Accelerating TensorFlow Lite on Qualcomm Hexagon DSPs](https:\u002F\u002Fblog.tensorflow.org\u002F2019\u002F12\u002Faccelerating-tensorflow-lite-on-qualcomm.html).\n* [Model Metadata](https:\u002F\u002Fwww.tensorflow.org\u002Flite\u002Fconvert\u002Fmetadata) - Provides a standard for model descriptions which also enables [Code Gen and Android Studio ML Model Binding](https:\u002F\u002Fwww.tensorflow.org\u002Flite\u002Finference_with_metadata\u002Fcodegen).\n\n## Models with samples\nHere are the TensorFlow Lite models with app \u002F device implementations, and references.\nNote: pretrained TensorFlow Lite models from MediaPipe are included, which you can implement with or without MediaPipe.\n\n### Computer vision\n\n#### Classification\n\n| Task | Model | App \\| Reference | Source |\n| ------------------------------- |-------------------------------------------------------------------------------------------------------------------------------------------------------------------| ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | -------------------|\n| Classification | MobileNetV1 ([download](https:\u002F\u002Fstorage.googleapis.com\u002Fdownload.tensorflow.org\u002Fmodels\u002Ftflite\u002Fmobilenet_v1_1.0_224_quant_and_labels.zip)) | [Android](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fexamples\u002Ftree\u002Fmaster\u002Flite\u002Fexamples\u002Fimage_classification\u002Fandroid) \\| [iOS](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fexamples\u002Ftree\u002Fmaster\u002Flite\u002Fexamples\u002Fimage_classification\u002Fios) \\| [Raspberry Pi](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fexamples\u002Ftree\u002Fmaster\u002Flite\u002Fexamples\u002Fimage_classification\u002Fraspberry_pi) \\| [Overview](https:\u002F\u002Fwww.tensorflow.org\u002Flite\u002Fmodels\u002Fimage_classification\u002Foverview) | tensorflow.org |\n| Classification | MobileNetV2 | Recognize Flowers on Android [Codelab](https:\u002F\u002Fcodelabs.developers.google.com\u002Fcodelabs\u002Frecognize-flowers-with-tensorflow-on-android\u002F#0) \\| [Android](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fexamples\u002Ftree\u002Fmaster\u002Flite\u002Fcodelabs\u002Fflower_classification\u002Fandroid) | TensorFlow team |\n| Classification | MobileNetV2 | Skin Lesion Detection [Android](https:\u002F\u002Fgithub.com\u002FAakashKumarNain\u002Fskin_cancer_detection\u002Ftree\u002Fmaster\u002Fdemo) | Community |\n| Classification | MobileNetV2 | American Sign Language Detection \\| [Colab Notebook](https:\u002F\u002Fcolab.research.google.com\u002Fdrive\u002F1xsunX7Qj_XWBZwcZLyjsKBg4RI0DNo2-?usp=sharing) \\| [Android](https:\u002F\u002Fgithub.com\u002Fsayannath\u002FAmerican-Sign-Language-Detection) | Community |\n| Classification | CNN + Quantisation Aware Training | Stone Paper Scissor Detection [Colab Notebook](https:\u002F\u002Fcolab.research.google.com\u002Fdrive\u002F1Wdso2N_76E8Xxniqd4C6T1sV5BuhKN1o?usp=sharing) \\| [Flutter](https:\u002F\u002Fgithub.com\u002Fsayannath\u002FAmerican-Sign-Language-Detection) | Community |\n| Classification | EfficientNet-Lite0 ([download](https:\u002F\u002Fgithub.com\u002Fmargaretmz\u002Ficon-classifier\u002Fblob\u002Fmaster\u002Fml-code\u002Ficons-50.tflite)) | Icon Classifier [Colab & Android](https:\u002F\u002Fgithub.com\u002Fmargaretmz\u002Ficon-classifier) \\| [tutorial 1](https:\u002F\u002Fmedium.com\u002Fswlh\u002Ficon-classifier-with-tflite-model-maker-9263c0021f72) \\| [tutorial 2](https:\u002F\u002Fmedium.com\u002F@margaretmz\u002Ficon-classifier-android-app-1fc0b727f761) | Community |\n\n#### Detection\n| Task | Model | App \\| Reference | Source |\n| -|-|-|-|\n| Object detection | Quantized COCO SSD MobileNet v1 ([download](https:\u002F\u002Fstorage.googleapis.com\u002Fdownload.tensorflow.org\u002Fmodels\u002Ftflite\u002Fcoco_ssd_mobilenet_v1_1.0_quant_2018_06_29.zip)) | [Android](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fexamples\u002Ftree\u002Fmaster\u002Flite\u002Fexamples\u002Fobject_detection\u002Fandroid) \\| [iOS](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fexamples\u002Ftree\u002Fmaster\u002Flite\u002Fexamples\u002Fobject_detection\u002Fios) \\| [Overview](https:\u002F\u002Fwww.tensorflow.org\u002Flite\u002Fmodels\u002Fobject_detection\u002Foverview#starter_model) | tensorflow.org |\n| Object detection | YOLO | [Flutter](https:\u002F\u002Fblog.francium.tech\u002Freal-time-object-detection-on-mobile-with-flutter-tensorflow-lite-and-yolo-android-part-a0042c9b62c6) \\| [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F1506.02640) | Community |\n| Object detection | [YOLOv5](https:\u002F\u002Ftfhub.dev\u002Fneso613\u002Flite-model\u002Fyolo-v5-tflite\u002Ftflite_model\u002F1) | [Yolov5 Inference ](https:\u002F\u002Fgithub.com\u002Fneso613\u002Fyolo-v5-tflite-model) | Community |\n| Object detection | MobileNetV2 SSD ([download](https:\u002F\u002Fgithub.com\u002Fgoogle\u002Fmediapipe\u002Ftree\u002Fmaster\u002Fmediapipe\u002Fmodels\u002Fssdlite_object_detection.tflite)) | [Reference](https:\u002F\u002Fgithub.com\u002Fgoogle\u002Fmediapipe\u002Fblob\u002Fmaster\u002Fmediapipe\u002Fmodels\u002Fobject_detection_saved_model\u002FREADME.md) | MediaPipe |\n| Object detection | MobileDet ([Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2004.14525)) | [Blog post (includes the TFLite conversion process)](https:\u002F\u002Fsayak.dev\u002Fmobiledet-optimization\u002F) | MobileDet is from University of Wisconsin-Madison and Google and the blog post is from the Community |\n| License Plate detection | SSD MobileNet [(download)](https:\u002F\u002Fgithub.com\u002FariG23498\u002FFlutter-License\u002Fblob\u002Fmaster\u002Fassets\u002Fdetect.tflite) | [Flutter](https:\u002F\u002Fgithub.com\u002FariG23498\u002FFlutter-License) | Community |\n| Face detection | BlazeFace ([download](https:\u002F\u002Fgithub.com\u002Fgoogle\u002Fmediapipe\u002Ftree\u002Fmaster\u002Fmediapipe\u002Fmodels\u002Fface_detection_front.tflite)) | [Paper](https:\u002F\u002Fsites.google.com\u002Fcorp\u002Fview\u002Fperception-cv4arvr\u002Fblazeface) | MediaPipe |\n| Face Authentication | [FaceNet](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1503.03832.pdf) | [Flutter](https:\u002F\u002Fgithub.com\u002Fsayannath\u002FFace-Authentication-App) | Community |\n| Hand detection & tracking | Palm detection & hand landmarks ([download](https:\u002F\u002Fgithub.com\u002Fgoogle\u002Fmediapipe\u002Ftree\u002Fmaster\u002Fmediapipe\u002Fmodels#hand-detection-and-tracking)) | [Blog post](https:\u002F\u002Fmediapipe.page.link\u002Fhandgoogleaiblog) \\| [Model card](https:\u002F\u002Fmediapipe.page.link\u002Fhandmc) \\| [Android](https:\u002F\u002Fgithub.com\u002Fsupremetech\u002Fmediapipe-demo-hand-detection) | MediaPipe & Community |\n\n#### Segmentation\n| Task | Model | App \\| Reference | Source |\n| -|-|-|-|\n| Segmentation | DeepLab V3 ([download](https:\u002F\u002Fstorage.googleapis.com\u002Fdownload.tensorflow.org\u002Fmodels\u002Ftflite\u002Fgpu\u002Fdeeplabv3_257_mv_gpu.tflite)) | [Android & iOS](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fexamples\u002Ftree\u002Fmaster\u002Flite\u002Fexamples\u002Fimage_segmentation\u002F) \\| [Overview](https:\u002F\u002Fwww.tensorflow.org\u002Flite\u002Fmodels\u002Fsegmentation\u002Foverview) \\| Flutter [Image](https:\u002F\u002Fgithub.com\u002Fkshitizrimal\u002FFlutter-TFLite-Image-Segmentation) \\| [Realtime](https:\u002F\u002Fgithub.com\u002Fkshitizrimal\u002Ftflite-realtime-flutter) \\| [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F1706.05587) | tf.org & Community |\n| Segmentation | Different variants of [DeepLab V3 models](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodels\u002Fblob\u002Fmaster\u002Fresearch\u002Fdeeplab\u002Fg3doc\u002Fmodel_zoo.md) | Models on [TF Hub](https:\u002F\u002Ftfhub.dev\u002Fs?module-type=image-segmentation&publisher=sayakpaul) with Colab Notebooks | Community |\n| Segmentation | [DeepLab V3 model](https:\u002F\u002Ftfhub.dev\u002Ftensorflow\u002Flite-model\u002Fdeeplabv3\u002F1\u002Fmetadata\u002F2?lite-format=tflite) | [Android](https:\u002F\u002Fgithub.com\u002Ffarmaker47\u002FUpdate_image_segmentation) \\| [Tutorial](https:\u002F\u002Ffarmaker47.medium.com\u002Fuse-camerax-with-image-segmentation-android-project-d8656f35cea3) | Community |\n| Hair Segmentation | [Download](https:\u002F\u002Fgithub.com\u002Fgoogle\u002Fmediapipe\u002Ftree\u002Fmaster\u002Fmediapipe\u002Fmodels\u002Fhair_segmentation.tflite) | [Paper](https:\u002F\u002Fsites.google.com\u002Fcorp\u002Fview\u002Fperception-cv4arvr\u002Fhair-segmentation) | MediaPipe |\n\n#### Style Transfer\n| Task | Model | App \\| Reference | Source |\n| -|-|-|-|\n| Style transfer | [Arbitrary image stylization](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmagenta\u002Ftree\u002Fmaster\u002Fmagenta\u002Fmodels\u002Farbitrary_image_stylization) | [Overview](https:\u002F\u002Fwww.tensorflow.org\u002Flite\u002Fmodels\u002Fstyle_transfer\u002Foverview) \\| [Android](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fexamples\u002Ftree\u002Fmaster\u002Flite\u002Fexamples\u002Fstyle_transfer\u002Fandroid) \\| [Flutter](https:\u002F\u002Fgithub.com\u002FPuzzleLeaf\u002Fflutter_tflite_style_transfer) | tf.org & Community |\n| Style transfer | Better-quality style transfer models in .tflite | Models on [TF Hub](https:\u002F\u002Ftfhub.dev\u002Fsayakpaul\u002Flite-model\u002Farbitrary-image-stylization-inceptionv3\u002Fdr\u002Fpredict\u002F1) with Colab Notebooks | Community |\n| Video Style Transfer | Download: \u003Cbr> [Dynamic range models](https:\u002F\u002Ftfhub.dev\u002Fsayakpaul\u002Flite-model\u002Farbitrary-image-stylization-inceptionv3-dynamic-shapes\u002Fdr\u002Ftransfer\u002F1)) | [Android](https:\u002F\u002Fgithub.com\u002Ffarmaker47\u002Fvideo_style_transfer) \\| [Tutorial](https:\u002F\u002Fmedium.com\u002F@farmaker47\u002Fandroid-implementation-of-video-style-transfer-with-tensorflow-lite-models-9338a6d2a3ea) | Community |\n| Segmentation & Style transfer | DeepLabV3 & Style Transfer [models](https:\u002F\u002Fgithub.com\u002Fmargaretmz\u002Fsegmentation-style-transfer\u002Ftree\u002Fmaster\u002Fml) | [Project repo](https:\u002F\u002Fgithub.com\u002Fmargaretmz\u002Fsegmentation-style-transfer) \\| [Android](https:\u002F\u002Fgithub.com\u002Fmargaretmz\u002Fsegmentation-style-transfer\u002Ftree\u002Fmaster\u002Fandroid) \\| [Tutorial](https:\u002F\u002Fmedium.com\u002Fgoogle-developer-experts\u002Fimage-background-stylizer-part-1-project-intro-d68c4547e7e3) | Community |\n#### Generative\n| Task | Model | App \\| Reference | Source |\n| -|-|-|-|\n| GANs | [U-GAT-IT](https:\u002F\u002Fgithub.com\u002Ftaki0112\u002FUGATIT) (Selfie2Anime) | [Project repo](https:\u002F\u002Fgithub.com\u002Fmargaretmz\u002Fselfie2anime-with-tflite) \\| [Android](https:\u002F\u002Fgithub.com\u002Fmargaretmz\u002Fselfie2anime-with-tflite\u002Ftree\u002Fmaster\u002Fandroid) \\| [Tutorial](https:\u002F\u002Fmedium.com\u002Fgoogle-developer-experts\u002Fselfie2anime-with-tflite-part-1-overview-f97500800ffe) | Community |\n| GANs | [White-box CartoonGAN](https:\u002F\u002Fgithub.com\u002FSystemErrorWang\u002FWhite-box-Cartoonization) ([download](https:\u002F\u002Ftfhub.dev\u002Fsayakpaul\u002Flite-model\u002Fcartoongan\u002Fdr\u002F1)) | [Project repo](https:\u002F\u002Fgithub.com\u002Fmargaretmz\u002FCartoonizer-with-TFLite) \\| [Android](https:\u002F\u002Fgithub.com\u002Fmargaretmz\u002FCartoonizer-with-TFLite\u002Ftree\u002Fmaster\u002Fandroid) \\| [Tutorial](https:\u002F\u002Fblog.tensorflow.org\u002F2020\u002F09\u002Fhow-to-create-cartoonizer-with-tf-lite.html) | Community |\n| GANs - Image Extrapolation | Boundless on [TF Hub](https:\u002F\u002Ftfhub.dev\u002Fsayakpaul\u002Flite-model\u002Fboundless-quarter\u002Fdr\u002F1) | [Colab Notebook](https:\u002F\u002Fcolab.research.google.com\u002Fgithub\u002Fsayakpaul\u002FAdventures-in-TensorFlow-Lite\u002Fblob\u002Fmaster\u002FBoundless_TFLite.ipynb) \\| [Original Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2003.06792v2.pdf) | Community |\n#### Post estimation\n| Task | Model | App \\| Reference | Source |\n| -|-|-|-|\n| Pose estimation | Posenet ([download](https:\u002F\u002Fstorage.googleapis.com\u002Fdownload.tensorflow.org\u002Fmodels\u002Ftflite\u002Fposenet_mobilenet_v1_100_257x257_multi_kpt_stripped.tflite)) | [Android](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fexamples\u002Ftree\u002Fmaster\u002Flite\u002Fexamples\u002Fposenet\u002Fandroid) \\| [iOS](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fexamples\u002Ftree\u002Fmaster\u002Flite\u002Fexamples\u002Fposenet\u002Fios) \\| [Overview](https:\u002F\u002Fwww.tensorflow.org\u002Flite\u002Fmodels\u002Fpose_estimation\u002Foverview) | tensorflow.org |\n| Pose Classification based Video Game Control | MoveNet Lightning ([download](https:\u002F\u002Fgithub.com\u002FNSTiwari\u002FVideo-Game-Control-using-Pose-Classification-and-TensorFlow-Lite\u002Fblob\u002Fmain\u002Fmovenet_lightning.tflite)) | [Project Repository](https:\u002F\u002Fgithub.com\u002FNSTiwari\u002FVideo-Game-Control-using-Pose-Classification-and-TensorFlow-Lite) | Community |\n\n\n#### Other\n| Task | Model | App \\| Reference | Source |\n| -|-|-|-|\n| Low-light image enhancement | [Models on TF Hub](https:\u002F\u002Ftfhub.dev\u002Fsayakpaul\u002Fmirnet-fixed\u002F1) | [Project repo](https:\u002F\u002Fgithub.com\u002Fsayakpaul\u002FMIRNet-TFLite) \\| [Original Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2003.06792v2.pdf) \\| [Flutter](https:\u002F\u002Fgithub.com\u002Fsayannath\u002FMIRNet-Flutter)| | Community |\n| OCR |[Models on TF Hub](https:\u002F\u002Ftfhub.dev\u002Ftulasiram58827\u002Flite-model\u002Fkeras-ocr\u002Fdr\u002F2) | [Project Repository](https:\u002F\u002Fgithub.com\u002Ftulasiram58827\u002Focr_tflite) | Community\n\n\n### Text\n| Task | Model | Sample apps | Source |\n| ------------------- |---------------------------------------------------------------------------------------------------------------------------------| ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------ |\n| Question & Answer | DistilBERT | [Android](https:\u002F\u002Fgithub.com\u002Fhuggingface\u002Ftflite-android-transformers\u002Fblob\u002Fmaster\u002Fbert) | Hugging Face |\n| Text Generation | GPT-2 \u002F DistilGPT2 | [Android](https:\u002F\u002Fgithub.com\u002Fhuggingface\u002Ftflite-android-transformers\u002Fblob\u002Fmaster\u002Fgpt2) | Hugging Face |\n| Text Classification | [Download](https:\u002F\u002Fstorage.googleapis.com\u002Fdownload.tensorflow.org\u002Fmodels\u002Ftflite\u002Ftext_classification\u002Ftext_classification.tflite) | [Android](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fexamples\u002Ftree\u002Fmaster\u002Flite\u002Fexamples\u002Ftext_classification\u002Fandroid) \\|[iOS](https:\u002F\u002Fgithub.com\u002Fkhurram18\u002FTextClassafication) \\| [Flutter](https:\u002F\u002Fgithub.com\u002Fam15h\u002Ftflite_flutter_plugin\u002Ftree\u002Fmaster\u002Fexample) | tf.org & Community |\n| Text Detection | CRAFT Text Detector ([Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1904.01941)) |[Download](https:\u002F\u002Fgithub.com\u002Ftulasiram58827\u002Fcraft_tflite\u002Fblob\u002Fmain\u002Fmodels\u002Fcraft_float_800.tflite?raw=true) \\| [Project Repository](https:\u002F\u002Fgithub.com\u002Ftulasiram58827\u002Fcraft_tflite\u002F) \\| [Blog1-Conversion to TFLite](https:\u002F\u002Ftulasi.dev\u002Fcraft-in-tflite) \\| [Blog2-EAST vs CRAFT](https:\u002F\u002Fsayak.dev\u002Foptimizing-text-detectors\u002F) \\| [Models on TF Hub](https:\u002F\u002Ftfhub.dev\u002Ftulasiram58827\u002Flite-model\u002Fcraft-text-detector\u002Fdr\u002F1) \\| Android (Coming Soon) | Community |\n| Text Detection | EAST Text Detector ([Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F1704.03155)) |[Models on TF Hub](https:\u002F\u002Ftfhub.dev\u002Fsayakpaul\u002Flite-model\u002Feast-text-detector\u002Fdr\u002F1) \\| [Conversion and Inference Notebook](https:\u002F\u002Fcolab.research.google.com\u002Fgithub\u002Fsayakpaul\u002FAdventures-in-TensorFlow-Lite\u002Fblob\u002Fmaster\u002FEAST_TFLite.ipynb) | Community |\n\n### Speech\n| Task | Model | App \\| Reference | Source |\n| ------------------ |------------------------------------| ------------------------------------------------------------------------------------- | ------------ |\n| Speech Recognition | DeepSpeech | [Reference](https:\u002F\u002Fgithub.com\u002Fmozilla\u002FDeepSpeech\u002Ftree\u002Fmaster\u002Fnative_client\u002Fjava) | Mozilla |\n| Speech Recognition | CONFORMER | [Inference](https:\u002F\u002Fgithub.com\u002Fneso613\u002FASR_TFLite) [Android](https:\u002F\u002Fgithub.com\u002Fwindmaple\u002Ftflite-asr) | Community |\n| Speech Synthesis | Tacotron-2, FastSpeech2, MB-Melgan | [Android](https:\u002F\u002Fgithub.com\u002FTensorSpeech\u002FTensorflowTTS\u002Ftree\u002Fmaster\u002Fexamples\u002Fandroid) | TensorSpeech |\n| Speech Synthesis(TTS) | Tacotron2, FastSpeech2, MelGAN, MB-MelGAN, HiFi-GAN, Parallel WaveGAN | [Inference Notebook](https:\u002F\u002Fgithub.com\u002Ftulasiram58827\u002FTTS_TFLite\u002Fblob\u002Fmain\u002FEnd_to_End_TTS.ipynb) \\| [Project Repository](https:\u002F\u002Fgithub.com\u002Ftulasiram58827\u002FTTS_TFLite\u002F) | Community |\n\n### Recommendation\n| Task | Model | App \\| Reference | Source |\n| ------------------ |------------------------------------| ------------------------------------------------------------------------------------- | ------------ |\n| On-device Recommendation | [Dual-Encoder](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fexamples\u002Ftree\u002Fmaster\u002Flite\u002Fexamples\u002Frecommendation\u002Fml) | [Android](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fexamples\u002Ftree\u002Fmaster\u002Flite\u002Fexamples\u002Frecommendation\u002Fandroid) \\| [iOS](https:\u002F\u002Fgithub.com\u002Fzhuzilin\u002Fon-device_recommendation_tflite) \\| [Reference](https:\u002F\u002Fblog.tensorflow.org\u002F2020\u002F09\u002Fintroduction-to-tflite-on-device-recommendation.html) | tf.org & Community |\n\n### Game\n| Task | Model | App \\| Reference | Source |\n| ------------------ |------------------------------------| ------------------------------------------------------------------------------------- | ------------ |\n| Game agent | Reinforcement learning | [Flutter](https:\u002F\u002Fgithub.com\u002Fwindmaple\u002Fplanestrike-flutter) \\| [Tutorial](https:\u002F\u002Fwindmaple.medium.com\u002F) | Community |\n\n\n\n## Model zoo\n\n### TensorFlow Lite models\nThese are the TensorFlow Lite models that could be implemented in apps and things:\n* [MobileNet](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodels\u002Fblob\u002Fmaster\u002Fresearch\u002Fslim\u002Fnets\u002Fmobilenet\u002FREADME.md) - Pretrained MobileNet v2 and v3 models.\n* TensorFlow Lite models\n * [TensorFlow Lite models](https:\u002F\u002Fwww.tensorflow.org\u002Flite\u002Fmodels) - With official Android and iOS examples.\n * [Pretrained models](https:\u002F\u002Fwww.tensorflow.org\u002Flite\u002Fguide\u002Fhosted_models) - Quantized and floating point variants.\n * [TensorFlow Hub](https:\u002F\u002Ftfhub.dev\u002F) - Set \"Model format = TFLite\" to find TensorFlow Lite models.\n\n### TensorFlow models\nThese are TensorFlow models that could be converted to .tflite and then implemented in apps and things:\n* [TensorFlow models](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodels\u002Ftree\u002Fmaster\u002Fofficial) - Official TensorFlow models.\n* [Tensorflow detection model zoo](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodels\u002Fblob\u002Fmaster\u002Fresearch\u002Fobject_detection\u002Fg3doc\u002Ftf2_detection_zoo.md) - Pre-trained on COCO, KITTI, AVA v2.1, iNaturalist Species datasets.\n\n## Ideas and Inspiration\n* [E2E TFLite Tutorials](https:\u002F\u002Fgithub.com\u002Fml-gde\u002Fe2e-tflite-tutorials) - Checkout this repo for sample app ideas and seeking help for your tutorial projects. Once a project gets completed, the links of the TensorFlow Lite model(s), sample code and tutorial will be added to this awesome list.\n\n## ML Kit examples\n[ML Kit](https:\u002F\u002Fdevelopers.google.com\u002Fml-kit) is a mobile SDK that brings Google's ML expertise to mobile developers.\n* 2019-10-01 [ML Kit Translate demo](https:\u002F\u002Fcodelabs.developers.google.com\u002Fcodelabs\u002Fmlkit-android-translate\u002F#0) - A tutorial with material design [Android](https:\u002F\u002Fgithub.com\u002Fgooglecodelabs\u002Fmlkit-android\u002Ftree\u002Fmaster\u002Ftranslate) (Kotlin) sample - recognize, identify Language and translate text from live camera with ML Kit for Firebase.\n* 2019-03-13 [Computer Vision with ML Kit - Flutter In Focus](https:\u002F\u002Fyoutu.be\u002FymyYUCrJnxU).\n* 2019-02-09 [Flutter + MLKit: Business Card Mail Extractor](https:\u002F\u002Fmedium.com\u002Fflutter-community\u002Fflutter-mlkit-8039ec66b6a) - A blog post with a [Flutter](https:\u002F\u002Fgithub.com\u002FDaemonLoki\u002FBusiness-Card-Mail-Extractor) sample code.\n* 2019-02-08 [From TensorFlow to ML Kit: Power your Android application with machine learning](https:\u002F\u002Fspeakerdeck.com\u002Fjinqian\u002Ffrom-tensorflow-to-ml-kit-power-your-android-application-with-machine-learning) - A talk with [Android](https:\u002F\u002Fgithub.com\u002Fxebia-france\u002Fmagritte) (Kotlin) sample code.\n* 2018-08-07 [Building a Custom Machine Learning Model on Android with TensorFlow Lite](https:\u002F\u002Fmedium.com\u002Fover-engineering\u002Fbuilding-a-custom-machine-learning-model-on-android-with-tensorflow-lite-26447e53abf2).\n* 2018-07-20 [ML Kit and Face Detection in Flutter](https:\u002F\u002Fflatteredwithflutter.com\u002Fml-kit-and-face-detection-in-flutter\u002F).\n* 2018-07-27 [ML Kit on Android 4: Landmark Detection](https:\u002F\u002Fmedium.com\u002Fgoogle-developer-experts\u002Fexploring-firebase-mlkit-on-android-landmark-detection-part-four-5e86b8deac3a).\n* 2018-07-28 [ML Kit on Android 3: Barcode Scanning](https:\u002F\u002Fmedium.com\u002Fgoogle-developer-experts\u002Fexploring-firebase-mlkit-on-android-barcode-scanning-part-three-cc6f5921a108).\n* 2018-05-31 [ML Kit on Android 2: Face Detection](https:\u002F\u002Fmedium.com\u002Fgoogle-developer-experts\u002Fexploring-firebase-mlkit-on-android-face-detection-part-two-de7e307c52e0).\n* 2018-05-22 [ML Kit on Android 1: Intro](https:\u002F\u002Fmedium.com\u002Fgoogle-developer-experts\u002Fexploring-firebase-mlkit-on-android-introducing-mlkit-part-one-98fcfedbeee0).\n\n## Plugins and SDKs\n* [Edge Impulse](https:\u002F\u002Fwww.edgeimpulse.com\u002F) - Created by [@EdgeImpulse](https:\u002F\u002Ftwitter.com\u002FEdgeImpulse) to help you to train TensorFlow Lite models for embedded devices in the cloud.\n* [MediaPipe](https:\u002F\u002Fgithub.com\u002Fgoogle\u002Fmediapipe) - A cross platform (mobile, desktop and Edge TPUs) AI pipeline by Google AI. (PM [Ming Yong](https:\u002F\u002Ftwitter.com\u002Frealmgyong)) | [MediaPipe examples](https:\u002F\u002Fmediapipe.readthedocs.io\u002Fen\u002Flatest\u002Fexamples.html).\n* [Coral Edge TPU](https:\u002F\u002Fcoral.ai\u002F) - Edge hardware by Google. [Coral Edge TPU examples](https:\u002F\u002Fcoral.ai\u002Fexamples\u002F).\n* [TensorFlow Lite Flutter Plugin](https:\u002F\u002Fgithub.com\u002Fam15h\u002Ftflite_flutter_plugin\u002F) - Provides a dart API similar to the TensorFlow Lite Java API for accessing TensorFlow Lite interpreter and performing inference in flutter apps. [tflite_flutter on pub.dev](https:\u002F\u002Fpub.dev\u002Fpackages\u002Ftflite_flutter).\n\n## Helpful links\n* [Netron](https:\u002F\u002Fgithub.com\u002Flutzroeder\u002Fnetron) - A tool for visualizing models.\n* [AI benchmark](http:\u002F\u002Fai-benchmark.com\u002Ftests.html) - A website for benchmarking computer vision models on smartphones.\n* [Performance measurement](https:\u002F\u002Fwww.tensorflow.org\u002Flite\u002Fperformance\u002Fmeasurement) - How to measure model performance on Android and iOS.\n* [Material design guidelines for ML](https:\u002F\u002Fmaterial.io\u002Fcollections\u002Fmachine-learning\u002Fpatterns-for-machine-learning-powered-features.html) - How to design machine learning powered features. A good example: [ML Kit Showcase App](https:\u002F\u002Fgithub.com\u002Ffirebase\u002Fmlkit-material-android).\n* [The People + AI Guide book](https:\u002F\u002Fpair.withgoogle.com\u002F) - Learn how to design human-centered AI products.\n* [Adventures in TensorFlow Lite](https:\u002F\u002Fgithub.com\u002Fsayakpaul\u002FAdventures-in-TensorFlow-Lite) - A repository showing non-trivial conversion processes and general explorations in TensorFlow Lite.\n* [TFProfiler](https:\u002F\u002Fgithub.com\u002Figlaweb\u002FTFProfiler) - An Android-based app to profile TensorFlow Lite models and measure its performance on smartphone.\n* [TensorFlow Lite for Microcontrollers](https:\u002F\u002Fwww.tensorflow.org\u002Flite\u002Fmicrocontrollers)\n* [TensorFlow Lite Examples - Android](https:\u002F\u002Fgithub.com\u002Fdailystudio\u002Ftensorflow-lite-examples-android) - A repository refactors and rewrites all the TensorFlow Lite Android examples which are included in the TensorFlow official website. \n* [Tensorflow-lite-kotlin-samples](https:\u002F\u002Fgithub.com\u002FSunitRoy2703\u002FTensorflow-lite-kotlin-samples) - A collection of Tensorflow Lite Android example Apps in Kotlin, to show different kinds of kotlin implementation of the [example apps](https:\u002F\u002Fwww.tensorflow.org\u002Flite\u002Fexamples)\n\n\n## Learning resources\nInterested but not sure how to get started? Here are some learning resources that will help you whether you are a beginner or a practitioner in the field for a while.\n\n### Blog posts\n\n* 2021-11-09 [On-device training in TensorFlow Lite](https:\u002F\u002Fblog.tensorflow.org\u002F2021\u002F11\u002Fon-device-training-in-tensorflow-lite.html)\n* 2021-09-27 [Optical character recognition with TensorFlow Lite: A new example app](https:\u002F\u002Fblog.tensorflow.org\u002F2021\u002F09\u002Fblog.tensorflow.org202109optical-character-recognition.html)\n* 2021-06-16 [https:\u002F\u002Fblog.tensorflow.org\u002F2021\u002F06\u002Feasier-object-detection-on-mobile-with-tf-lite.html](https:\u002F\u002Fblog.tensorflow.org\u002F2021\u002F11\u002Fon-device-training-in-tensorflow-lite.html)\n* 2020-12-29 [YOLOv3 to TensorFlow Lite Conversion](https:\u002F\u002Fmedium.com\u002Fanalytics-vidhya\u002Fyolov3-to-tensorflow-lite-conversion-4602cec5c239) - By Nitin Tiwari.\n* 2020-04-20 [What is new in TensorFlow Lite](https:\u002F\u002Fblog.tensorflow.org\u002F2020\u002F04\u002Fwhats-new-in-tensorflow-lite-from-devsummit-2020.html) - By Khanh LeViet.\n* 2020-04-17 [Optimizing style transfer to run on mobile with TFLite](https:\u002F\u002Fblog.tensorflow.org\u002F2020\u002F04\u002Foptimizing-style-transfer-to-run-on-mobile-with-tflite.html) - By Khanh LeViet and Luiz Gustavo Martins.\n* 2020-04-14 [How TensorFlow Lite helps you from prototype to product](https:\u002F\u002Fblog.tensorflow.org\u002F2020\u002F04\u002Fhow-tensorflow-lite-helps-you-from-prototype-to-product.html) - By Khanh LeViet.\n* 2019-11-08 [Getting Started with ML on MCUs with TensorFlow](https:\u002F\u002Fblog.particle.io\u002F2019\u002F11\u002F08\u002Fparticle-machine-learning-101\u002F) - By Brandon Satrom.\n* 2019-08-05 [TensorFlow Model Optimization Toolkit — float16 quantization halves model size](https:\u002F\u002Fblog.tensorflow.org\u002F2019\u002F08\u002Ftensorflow-model-optimization-toolkit_5.html) - By the TensorFlow team.\n* 2018-07-13 [Training and serving a real-time mobile object detector in 30 minutes with Cloud TPUs](https:\u002F\u002Fblog.tensorflow.org\u002F2018\u002F07\u002Ftraining-and-serving-realtime-mobile-object-detector-cloud-tpus.html) - By Sara Robinson, Aakanksha Chowdhery, and Jonathan Huang.\n* 2018-06-11 - [Why the Future of Machine Learning is Tiny](https:\u002F\u002Fpetewarden.com\u002F2018\u002F06\u002F11\u002Fwhy-the-future-of-machine-learning-is-tiny\u002F) - By Pete Warden.\n* 2018-03-30 - [Using TensorFlow Lite on Android](https:\u002F\u002Fblog.tensorflow.org\u002F2018\u002F03\u002Fusing-tensorflow-lite-on-android.html)) - By Laurence Moroney.\n\n### Books\n* 2021-12-01 [AI and Machine Learning On-Device Development](https:\u002F\u002Flearning.oreilly.com\u002Flibrary\u002Fview\u002Fai-and-machine\u002F9781098101732\u002F) (early access) - By Laurence Moroney ([@lmoroney](https:\u002F\u002Ftwitter.com\u002Flmoroney)).\n* 2020-10-01 [AI and Machine Learning for Coders](https:\u002F\u002Flearning.oreilly.com\u002Flibrary\u002Fview\u002Fai-and-machine\u002F9781492078180\u002F) - By Laurence Moroney ([@lmoroney](https:\u002F\u002Ftwitter.com\u002Flmoroney)).\n* 2020-04-06 [Mobile Deep Learning with TensorFlow Lite, ML Kit and Flutter](https:\u002F\u002Fwww.packtpub.com\u002Fproduct\u002Fmobile-deep-learning-with-tensorflow-lite-ml-kit-and-flutter\u002F9781789611212): Build scalable real-world projects to implement end-to-end neural networks on Android and iOS ([GitHub](https:\u002F\u002Fgithub.com\u002FPacktPublishing\u002FMobile-Deep-Learning-Projects)) - By Anubhav Singh ([@xprilion](https:\u002F\u002Fgithub.com\u002Fxprilion)) and Rimjhim Bhadani ([@Rimjhim28](https:\u002F\u002Fgithub.com\u002FRimjhim28)).\n* 2020-03-01 Raspberry Pi for Computer Vision ([Complete Bundle](https:\u002F\u002Fwww.pyimagesearch.com\u002Fraspberry-pi-for-computer-vision) | [TOC](https:\u002F\u002Fwww.pyimagesearch.com\u002F2019\u002F04\u002F05\u002Ftable-of-contents-raspberry-pi-for-computer-vision\u002F)) - By the PyImageSearch Team: Adrian Rosebrock ([@PyImageSearch](https:\u002F\u002Ftwitter.com\u002FPyImageSearch)), David Hoffman, Asbhishek Thanki, Sayak Paul ([@RisingSayak](https:\u002F\u002Ftwitter.com\u002FRisingSayak)), and David Mcduffee.\n* 2019-12-01 [TinyML](http:\u002F\u002Fshop.oreilly.com\u002Fproduct\u002F0636920254508.do) - By Pete Warden ([@petewarden](https:\u002F\u002Ftwitter.com\u002Fpetewarden)) and Daniel Situnayake ([@dansitu](https:\u002F\u002Ftwitter.com\u002Fdansitu)).\n* 2019-10-01 [Practical Deep Learning for Cloud, Mobile, and Edge](https:\u002F\u002Fwww.practicaldeeplearning.ai\u002F) - By Anirudh Koul ([@AnirudhKoul](https:\u002F\u002Ftwitter.com\u002FAnirudhKoul)), Siddha Ganju ([@SiddhaGanju](https:\u002F\u002Ftwitter.com\u002FSiddhaGanju)), and Meher Kasam ([@MeherKasam](https:\u002F\u002Ftwitter.com\u002FMeherKasam)).\n\n### Videos\n* 2021-10-06 [Contributing to TensorFlow Lite with Sunit Roy](https:\u002F\u002Fyoutu.be\u002FsZayUoWW6nE) (Hacktoberfest 2021)\n* 2020-07-25 [Android ML by Hoi Lam](https:\u002F\u002Fyoutu.be\u002Fm_bEh8YifnQ) (GDG Kolkata meetup).\n* 2020-04-01 [Easy on-device ML from prototype to production](https:\u002F\u002Fyoutu.be\u002FALxWJoh_BHw) (TF Dev Summit 2020).\n* 2020-03-11 [TensorFlow Lite: ML for mobile and IoT devices](https:\u002F\u002Fyoutu.be\u002F27Zx-4GOQA8) (TF Dev Summit 2020).\n* 2019-10-31 [Keynote - TensorFlow Lite: ML for mobile and IoT devices](https:\u002F\u002Fyoutu.be\u002FzjDGAiLqGk8).\n* 2019-10-31 [TensorFlow Lite: Solution for running ML on-device](https:\u002F\u002Fyoutu.be\u002F0SpZy7iouFU).\n* 2019-10-31 [TensorFlow model optimization: Quantization and pruning](https:\u002F\u002Fyoutu.be\u002F3JWRVx1OKQQ).\n* 2019-10-29 [Inside TensorFlow: TensorFlow Lite](https:\u002F\u002Fyoutu.be\u002FgHN0jDbJz8E).\n* 2018-04-18 [TensorFlow Lite for Android (Coding TensorFlow)](https:\u002F\u002Fyoutu.be\u002FJnhW5tQ_7Vo).\n\n### Podcasts\n* 2020-08-08 [Talking Machine Learning with Hoi Lam](https:\u002F\u002Fanchor.fm\u002Ftalkingwithapples\u002Fepisodes\u002FTalking-Machine-Learning-with-Hoi-Lam-eiaj7v).\n\n### MOOCs\n* [Introduction to TensorFlow Lite](https:\u002F\u002Fwww.udacity.com\u002Fcourse\u002Fintro-to-tensorflow-lite--ud190) - Udacity course by Daniel Situnayake (@dansitu), Paige Bailey ([@DynamicWebPaige](https:\u002F\u002Ftwitter.com\u002FDynamicWebPaige)), and Juan Delgado.\n* [Device-based Models with TensorFlow Lite](https:\u002F\u002Fwww.coursera.org\u002Flearn\u002Fdevice-based-models-tensorflow) - Coursera course by Laurence Moroney ([@lmoroney](https:\u002F\u002Ftwitter.com\u002Flmoroney)).\n* [The Future of ML is Tiny and Bright](https:\u002F\u002Fwww.edx.org\u002Fprofessional-certificate\u002Fharvardx-tiny-machine-learning) - A series of edX courses created by Harvard in collaboration with Google. Instructors - Vijay Janapa Reddi, Laurence Moroney, and Pete Warden.\n","\u003Cp align=\"center\">\n \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmargaretmz_awesome-tensorflow-lite_readme_d02815ab1b45.png\" alt=\"awesome tflite\" width=\"500\">\n\u003C\u002Fp>\n\n\u003C!-- omit in toc -->\n# 令人惊叹的 TensorFlow Lite [![Awesome](https:\u002F\u002Fawesome.re\u002Fbadge.svg)](https:\u002F\u002Fawesome.re) [![欢迎 PR](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FPRs-welcome-brightgreen.svg?style=flat-square)](http:\u002F\u002Fmakeapullrequest.com) [![Twitter](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FTwitter-%40margaretmz-blue)](https:\u002F\u002Ftwitter.com\u002Fmargaretmz)\n\n[TensorFlow Lite](https:\u002F\u002Fwww.tensorflow.org\u002Flite) 是一套工具，可帮助将 TensorFlow 模型转换并优化，以便在移动设备和边缘设备上运行。目前已有超过 40 亿台设备在使用它！借助 TensorFlow 2.x，你可以用 tf.Keras 训练模型，轻松将其转换为 .tflite 格式并部署；或者直接从模型库中下载预训练的 TensorFlow Lite 模型。\n\n这是一个包含 TensorFlow Lite 模型、示例应用、实用工具和学习资源的精选列表——\n* 展示社区利用 TensorFlow Lite 所构建的各种成果\n* 将所有示例并排展示，方便参考\n* 分享知识与学习资源\n\n如果你希望贡献内容，请提交 PR，并遵循[此处](CONTRIBUTING.md)的指南。\n\n\u003C!-- omit in toc -->\n ## 目录\n- [过往公告：](#past-announcements)\n- [附带示例的模型](#models-with-samples)\n - [计算机视觉](#computer-vision)\n - [分类](#classification)\n - [检测](#detection)\n - [分割](#segmentation)\n - [风格迁移](#style-transfer)\n - [生成式](#generative)\n - [后处理估计](#post-estimation)\n - [其他](#other)\n - [文本](#text)\n - [语音](#speech)\n - [推荐](#recommendation)\n - [游戏](#game)\n- [模型库](#model-zoo)\n - [TensorFlow Lite 模型](#tensorflow-lite-models)\n - [TensorFlow 模型](#tensorflow-models)\n- [创意与灵感](#ideas-and-inspiration)\n- [ML Kit 示例](#ml-kit-examples)\n- [插件与 SDK](#plugins-and-sdks)\n- [实用链接](#helpful-links)\n- [学习资源](#learning-resources)\n - [博客文章](#blog-posts)\n - [书籍](#books)\n - [视频](#videos)\n - [播客](#podcasts)\n - [慕课](#moocs)\n\n## 过往公告：\n以下是 TensorFlow Lite 的一些重要功能发布信息：\n* [新转换器发布公告](https:\u002F\u002Fgroups.google.com\u002Fa\u002Ftensorflow.org\u002Fd\u002Fmsg\u002Ftflite\u002FZ_h7706dt8Q\u002FsNrjPj4yGgAJ) — 基于 [MLIR](https:\u002F\u002Fmedium.com\u002Ftensorflow\u002Fmlir-a-new-intermediate-representation-and-compiler-framework-beba999ed18d) 的新转换器，支持转换 Mask R-CNN 和 Mobile BERT 等新型模型，同时增强了对函数式控制流的支持以及转换过程中的错误处理能力。该功能已在 nightly 版本中默认启用。\n* [Android 支持库](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Ftflite-support\u002Ftree\u002Fmaster\u002Ftensorflow_lite_support\u002Fjava) — 让移动端开发更加便捷（参见 [Android](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fexamples\u002Fblob\u002Fmaster\u002Flite\u002Fexamples\u002Fimage_classification\u002Fandroid\u002FEXPLORE_THE_CODE.md) 示例代码）。\n* [模型生成器](https:\u002F\u002Fwww.tensorflow.org\u002Flite\u002Fguide\u002Fmodel_maker) — 仅需几行代码即可轻松创建自定义的图像和文本分类模型。下方提供了由社区制作的图标分类器教程。\n* [设备端训练](https:\u002F\u002Fblog.tensorflow.org\u002F2019\u002F12\u002Fexample-on-device-model-personalization.html) — 终于来了！目前仅限于图像分类任务上的迁移学习，但已是一个良好的开端。请参阅官方的 [Android](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fexamples\u002Fblob\u002Fmaster\u002Flite\u002Fexamples\u002Fmodel_personalization\u002FREADME.md) 示例代码，以及另一份来自社区的实现（[博客](https:\u002F\u002Faqibsaeed.github.io\u002Fon-device-activity-recognition) | [Android](https:\u002F\u002Fgithub.com\u002Faqibsaeed\u002Fon-device-activity-recognition)）。\n* [Hexagon 委托](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Ftensorflow\u002Fblob\u002Fmaster\u002Ftensorflow\u002Flite\u002Fg3doc\u002Fperformance\u002Fhexagon_delegate.md) — 如何使用 Hexagon 委托加速移动和边缘设备上的模型推理。另请参阅博文 [在高通 Hexagon DSP 上加速 TensorFlow Lite](https:\u002F\u002Fblog.tensorflow.org\u002F2019\u002F12\u002Faccelerating-tensorflow-lite-on-qualcomm.html)。\n* [模型元数据](https:\u002F\u002Fwww.tensorflow.org\u002Flite\u002Fconvert\u002Fmetadata) — 提供了一套标准化的模型描述格式，同时还支持 [代码生成和 Android Studio ML 模型绑定](https:\u002F\u002Fwww.tensorflow.org\u002Flite\u002Finference_with_metadata\u002Fcodegen)。\n\n## 附带示例的模型\n以下列出了带有应用程序或设备实现的 TensorFlow Lite 模型及其相关参考资料。\n注意：其中包含了来自 MediaPipe 的预训练 TensorFlow Lite 模型，你既可以结合 MediaPipe 使用，也可以独立实现。\n\n### 计算机视觉\n\n#### 分类\n\n| 任务 | 模型 | 应用 \\| 参考 | 来源 |\n| ------------------------------- |-------------------------------------------------------------------------------------------------------------------------------------------------------------------| ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | -------------------|\n| 分类 | MobileNetV1 ([下载](https:\u002F\u002Fstorage.googleapis.com\u002Fdownload.tensorflow.org\u002Fmodels\u002Ftflite\u002Fmobilenet_v1_1.0_224_quant_and_labels.zip)) | [Android](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fexamples\u002Ftree\u002Fmaster\u002Flite\u002Fexamples\u002Fimage_classification\u002Fandroid) \\| [iOS](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fexamples\u002Ftree\u002Fmaster\u002Flite\u002Fexamples\u002Fimage_classification\u002Fios) \\| [树莓派](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fexamples\u002Ftree\u002Fmaster\u002Flite\u002Fexamples\u002Fimage_classification\u002Fraspberry_pi) \\| [概览](https:\u002F\u002Fwww.tensorflow.org\u002Flite\u002Fmodels\u002Fimage_classification\u002Foverview) | tensorflow.org |\n| 分类 | MobileNetV2 | 在 Android 上识别花卉 [Codelab](https:\u002F\u002Fcodelabs.developers.google.com\u002Fcodelabs\u002Frecognize-flowers-with-tensorflow-on-android\u002F#0) \\| [Android](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fexamples\u002Ftree\u002Fmaster\u002Flite\u002Fcodelabs\u002Fflower_classification\u002Fandroid) | TensorFlow 团队 |\n| 分类 | MobileNetV2 | 皮肤病变检测 [Android](https:\u002F\u002Fgithub.com\u002FAakashKumarNain\u002Fskin_cancer_detection\u002Ftree\u002Fmaster\u002Fdemo) | 社区 |\n| 分类 | MobileNetV2 | 美国手语识别 \\| [Colab 笔记本](https:\u002F\u002Fcolab.research.google.com\u002Fdrive\u002F1xsunX7Qj_XWBZwcZLyjsKBg4RI0DNo2-?usp=sharing) \\| [Android](https:\u002F\u002Fgithub.com\u002Fsayannath\u002FAmerican-Sign-Language-Detection) | 社区 |\n| 分类 | CNN + 量化感知训练 | 石头剪刀布识别 [Colab 笔记本](https:\u002F\u002Fcolab.research.google.com\u002Fdrive\u002F1Wdso2N_76E8Xxniqd4C6T1sV5BuhKN1o?usp=sharing) \\| [Flutter](https:\u002F\u002Fgithub.com\u002Fsayannath\u002FAmerican-Sign-Language-Detection) | 社区 |\n| 分类 | EfficientNet-Lite0 ([下载](https:\u002F\u002Fgithub.com\u002Fmargaretmz\u002Ficon-classifier\u002Fblob\u002Fmaster\u002Fml-code\u002Ficons-50.tflite)) | 图标分类器 [Colab & Android](https:\u002F\u002Fgithub.com\u002Fmargaretmz\u002Ficon-classifier) \\| [教程 1](https:\u002F\u002Fmedium.com\u002Fswlh\u002Ficon-classifier-with-tflite-model-maker-9263c0021f72) \\| [教程 2](https:\u002F\u002Fmedium.com\u002F@margaretmz\u002Ficon-classifier-android-app-1fc0b727f761) | 社区 |\n\n#### 检测\n| 任务 | 模型 | 应用 \\| 参考 | 来源 |\n| -|-|-|-|\n| 目标检测 | 量化后的 COCO SSD MobileNet v1（[下载](https:\u002F\u002Fstorage.googleapis.com\u002Fdownload.tensorflow.org\u002Fmodels\u002Ftflite\u002Fcoco_ssd_mobilenet_v1_1.0_quant_2018_06_29.zip)） | [Android](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fexamples\u002Ftree\u002Fmaster\u002Flite\u002Fexamples\u002Fobject_detection\u002Fandroid) \\| [iOS](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fexamples\u002Ftree\u002Fmaster\u002Flite\u002Fexamples\u002Fobject_detection\u002Fios) \\| [概述](https:\u002F\u002Fwww.tensorflow.org\u002Flite\u002Fmodels\u002Fobject_detection\u002Foverview#starter_model) | tensorflow.org |\n| 目标检测 | YOLO | [Flutter](https:\u002F\u002Fblog.francium.tech\u002Freal-time-object-detection-on-mobile-with-flutter-tensorflow-lite-and-yolo-android-part-a0042c9b62c6) \\| [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F1506.02640) | 社区 |\n| 目标检测 | [YOLOv5](https:\u002F\u002Ftfhub.dev\u002Fneso613\u002Flite-model\u002Fyolo-v5-tflite\u002Ftflite_model\u002F1) | [Yolov5 推理](https:\u002F\u002Fgithub.com\u002Fneso613\u002Fyolo-v5-tflite-model) | 社区 |\n| 目标检测 | MobileNetV2 SSD（[下载](https:\u002F\u002Fgithub.com\u002Fgoogle\u002Fmediapipe\u002Ftree\u002Fmaster\u002Fmediapipe\u002Fmodels\u002Fssdlite_object_detection.tflite)) | [参考](https:\u002F\u002Fgithub.com\u002Fgoogle\u002Fmediapipe\u002Fblob\u002Fmaster\u002Fmediapipe\u002Fmodels\u002Fobject_detection_saved_model\u002FREADME.md) | MediaPipe |\n| 目标检测 | MobileDet（[论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2004.14525)） | [博客文章（包含 TFLite 转换过程）](https:\u002F\u002Fsayak.dev\u002Fmobiledet-optimization\u002F) | MobileDet 来自威斯康星大学麦迪逊分校和谷歌，博客文章由社区发布 |\n| 车牌识别 | SSD MobileNet [(下载)](https:\u002F\u002Fgithub.com\u002FariG23498\u002FFlutter-License\u002Fblob\u002Fmaster\u002Fassets\u002Fdetect.tflite) | [Flutter](https:\u002F\u002Fgithub.com\u002FariG23498\u002FFlutter-License) | 社区 |\n| 人脸检测 | BlazeFace（[下载](https:\u002F\u002Fgithub.com\u002Fgoogle\u002Fmediapipe\u002Ftree\u002Fmaster\u002Fmediapipe\u002Fmodels\u002Fface_detection_front.tflite)) | [论文](https:\u002F\u002Fsites.google.com\u002Fcorp\u002Fview\u002Fperception-cv4arvr\u002Fblazeface) | MediaPipe |\n| 人脸识别认证 | [FaceNet](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1503.03832.pdf) | [Flutter](https:\u002F\u002Fgithub.com\u002Fsayannath\u002FFace-Authentication-App) | 社区 |\n| 手部检测与跟踪 | 掌部检测与手部关键点（[下载](https:\u002F\u002Fgithub.com\u002Fgoogle\u002Fmediapipe\u002Ftree\u002Fmaster\u002Fmediapipe\u002Fmodels#hand-detection-and-tracking)) | [博客文章](https:\u002F\u002Fmediapipe.page.link\u002Fhandgoogleaiblog) \\| [模型卡片](https:\u002F\u002Fmediapipe.page.link\u002Fhandmc) \\| [Android](https:\u002F\u002Fgithub.com\u002Fsupremetech\u002Fmediapipe-demo-hand-detection) | MediaPipe & 社区 |\n\n#### 分割\n| 任务 | 模型 | 应用 \\| 参考 | 来源 |\n| -|-|-|-|\n| 分割 | DeepLab V3 ([下载](https:\u002F\u002Fstorage.googleapis.com\u002Fdownload.tensorflow.org\u002Fmodels\u002Ftflite\u002Fgpu\u002Fdeeplabv3_257_mv_gpu.tflite)) | [Android & iOS](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fexamples\u002Ftree\u002Fmaster\u002Flite\u002Fexamples\u002Fimage_segmentation\u002F) \\| [概述](https:\u002F\u002Fwww.tensorflow.org\u002Flite\u002Fmodels\u002Fsegmentation\u002Foverview) \\| Flutter [图像](https:\u002F\u002Fgithub.com\u002Fkshitizrimal\u002FFlutter-TFLite-Image-Segmentation) \\| [实时](https:\u002F\u002Fgithub.com\u002Fkshitizrimal\u002Ftflite-realtime-flutter) \\| [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F1706.05587) | tf.org & 社区 |\n| 分割 | [DeepLab V3 模型](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodels\u002Fblob\u002Fmaster\u002Fresearch\u002Fdeeplab\u002Fg3doc\u002Fmodel_zoo.md)的不同变体 | [TF Hub](https:\u002F\u002Ftfhub.dev\u002Fs?module-type=image-segmentation&publisher=sayakpaul)上的模型，附带Colab笔记本 | 社区 |\n| 分割 | [DeepLab V3 模型](https:\u002F\u002Ftfhub.dev\u002Ftensorflow\u002Flite-model\u002Fdeeplabv3\u002F1\u002Fmetadata\u002F2?lite-format=tflite) | [Android](https:\u002F\u002Fgithub.com\u002Ffarmaker47\u002FUpdate_image_segmentation) \\| [教程](https:\u002F\u002Ffarmaker47.medium.com\u002Fuse-camerax-with-image-segmentation-android-project-d8656f35cea3) | 社区 |\n| 头发分割 | [下载](https:\u002F\u002Fgithub.com\u002Fgoogle\u002Fmediapipe\u002Ftree\u002Fmaster\u002Fmediapipe\u002Fmodels\u002Fhair_segmentation.tflite) | [论文](https:\u002F\u002Fsites.google.com\u002Fcorp\u002Fview\u002Fperception-cv4arvr\u002Fhair-segmentation) | MediaPipe |\n\n#### 风格迁移\n| 任务 | 模型 | 应用 \\| 参考 | 来源 |\n| -|-|-|-|\n| 风格迁移 | [任意图像风格化](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmagenta\u002Ftree\u002Fmaster\u002Fmagenta\u002Fmodels\u002Farbitrary_image_stylization) | [概览](https:\u002F\u002Fwww.tensorflow.org\u002Flite\u002Fmodels\u002Fstyle_transfer\u002Foverview) \\| [Android](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fexamples\u002Ftree\u002Fmaster\u002Flite\u002Fexamples\u002Fstyle_transfer\u002Fandroid) \\| [Flutter](https:\u002F\u002Fgithub.com\u002FPuzzleLeaf\u002Fflutter_tflite_style_transfer) | tf.org & 社区 |\n| 风格迁移 | 更高质量的 .tflite 格式风格迁移模型 | [TF Hub](https:\u002F\u002Ftfhub.dev\u002Fsayakpaul\u002Flite-model\u002Farbitrary-image-stylization-inceptionv3\u002Fdr\u002Fpredict\u002F1) 上的模型，附 Colab 笔记本 | 社区 |\n| 视频风格迁移 | 下载： \u003Cbr> [动态范围模型](https:\u002F\u002Ftfhub.dev\u002Fsayakpaul\u002Flite-model\u002Farbitrary-image-stylization-inceptionv3-dynamic-shapes\u002Fdr\u002Ftransfer\u002F1)) | [Android](https:\u002F\u002Fgithub.com\u002Ffarmaker47\u002Fvideo_style_transfer) \\| [教程](https:\u002F\u002Fmedium.com\u002F@farmaker47\u002Fandroid-implementation-of-video-style-transfer-with-tensorflow-lite-models-9338a6d2a3ea) | 社区 |\n| 分割与风格迁移 | DeepLabV3 和风格迁移 [模型](https:\u002F\u002Fgithub.com\u002Fmargaretmz\u002Fsegmentation-style-transfer\u002Ftree\u002Fmaster\u002Fml) | [项目仓库](https:\u002F\u002Fgithub.com\u002Fmargaretmz\u002Fsegmentation-style-transfer) \\| [Android](https:\u002F\u002Fgithub.com\u002Fmargaretmz\u002Fsegmentation-style-transfer\u002Ftree\u002Fmaster\u002Fandroid) \\| [教程](https:\u002F\u002Fmedium.com\u002Fgoogle-developer-experts\u002Fimage-background-stylizer-part-1-project-intro-d68c4547e7e3) | 社区 |\n#### 生成模型\n| 任务 | 模型 | 应用 \\| 参考 | 来源 |\n| -|-|-|-|\n| GANs | [U-GAT-IT](https:\u002F\u002Fgithub.com\u002Ftaki0112\u002FUGATIT) (自拍转动漫) | [项目仓库](https:\u002F\u002Fgithub.com\u002Fmargaretmz\u002Fselfie2anime-with-tflite) \\| [Android](https:\u002F\u002Fgithub.com\u002Fmargaretmz\u002Fselfie2anime-with-tflite\u002Ftree\u002Fmaster\u002Fandroid) \\| [教程](https:\u002F\u002Fmedium.com\u002Fgoogle-developer-experts\u002Fselfie2anime-with-tflite-part-1-overview-f97500800ffe) | 社区 |\n| GANs | [白盒卡通化GAN](https:\u002F\u002Fgithub.com\u002FSystemErrorWang\u002FWhite-box-Cartoonization) ([下载](https:\u002F\u002Ftfhub.dev\u002Fsayakpaul\u002Flite-model\u002Fcartoongan\u002Fdr\u002F1)) | [项目仓库](https:\u002F\u002Fgithub.com\u002Fmargaretmz\u002FCartoonizer-with-TFLite) \\| [Android](https:\u002F\u002Fgithub.com\u002Fmargaretmz\u002FCartoonizer-with-TFLite\u002Ftree\u002Fmaster\u002Fandroid) \\| [教程](https:\u002F\u002Fblog.tensorflow.org\u002F2020\u002F09\u002Fhow-to-create-cartoonizer-with-tf-lite.html) | 社区 |\n| GANs - 图像外推 | Boundless 在 [TF Hub](https:\u002F\u002Ftfhub.dev\u002Fsayakpaul\u002Flite-model\u002Fboundless-quarter\u002Fdr\u002F1) 上 | [Colab 笔记本](https:\u002F\u002Fcolab.research.google.com\u002Fgithub\u002Fsayakpaul\u002FAdventures-in-TensorFlow-Lite\u002Fblob\u002Fmaster\u002FBoundless_TFLite.ipynb) \\| [原始论文](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2003.06792v2.pdf) | 社区 |\n#### 姿态估计后处理\n| 任务 | 模型 | 应用 \\| 参考 | 来源 |\n| -|-|-|-|\n| 姿态估计 | Posenet ([下载](https:\u002F\u002Fstorage.googleapis.com\u002Fdownload.tensorflow.org\u002Fmodels\u002Ftflite\u002Fposenet_mobilenet_v1_100_257x257_multi_kpt_stripped.tflite)) | [Android](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fexamples\u002Ftree\u002Fmaster\u002Flite\u002Fexamples\u002Fposenet\u002Fandroid) \\| [iOS](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fexamples\u002Ftree\u002Fmaster\u002Flite\u002Fexamples\u002Fposenet\u002Fios) \\| [概览](https:\u002F\u002Fwww.tensorflow.org\u002Flite\u002Fmodels\u002Fpose_estimation\u002Foverview) | tensorflow.org |\n| 基于姿态分类的视频游戏控制 | MoveNet Lightning ([下载](https:\u002F\u002Fgithub.com\u002FNSTiwari\u002FVideo-Game-Control-using-Pose-Classification-and-TensorFlow-Lite\u002Fblob\u002Fmain\u002Fmovenet_lightning.tflite)) | [项目仓库](https:\u002F\u002Fgithub.com\u002FNSTiwari\u002FVideo-Game-Control-using-Pose-Classification-and-TensorFlow-Lite) | 社区 |\n\n\n#### 其他\n| 任务 | 模型 | 应用 \\| 参考 | 来源 |\n| -|-|-|-|\n| 低光图像增强 | [TF Hub 上的模型](https:\u002F\u002Ftfhub.dev\u002Fsayakpaul\u002Fmirnet-fixed\u002F1) | [项目仓库](https:\u002F\u002Fgithub.com\u002Fsayakpaul\u002FMIRNet-TFLite) \\| [原始论文](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2003.06792v2.pdf) \\| [Flutter](https:\u002F\u002Fgithub.com\u002Fsayannath\u002FMIRNet-Flutter)| | 社区 |\n| OCR |[TF Hub 上的模型](https:\u002F\u002Ftfhub.dev\u002Ftulasiram58827\u002Flite-model\u002Fkeras-ocr\u002Fdr\u002F2) | [项目仓库](https:\u002F\u002Fgithub.com\u002Ftulasiram58827\u002Focr_tflite) | 社区\n\n### 文本\n| 任务 | 模型 | 示例应用 | 来源 |\n| ------------------- |---------------------------------------------------------------------------------------------------------------------------------| ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------ |\n| 问答 | DistilBERT | [Android](https:\u002F\u002Fgithub.com\u002Fhuggingface\u002Ftflite-android-transformers\u002Fblob\u002Fmaster\u002Fbert) | Hugging Face |\n| 文本生成 | GPT-2 \u002F DistilGPT2 | [Android](https:\u002F\u002Fgithub.com\u002Fhuggingface\u002Ftflite-android-transformers\u002Fblob\u002Fmaster\u002Fgpt2) | Hugging Face |\n| 文本分类 | [下载](https:\u002F\u002Fstorage.googleapis.com\u002Fdownload.tensorflow.org\u002Fmodels\u002Ftflite\u002Ftext_classification\u002Ftext_classification.tflite) | [Android](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fexamples\u002Ftree\u002Fmaster\u002Flite\u002Fexamples\u002Ftext_classification\u002Fandroid) \\|[iOS](https:\u002F\u002Fgithub.com\u002Fkhurram18\u002FTextClassafication) \\| [Flutter](https:\u002F\u002Fgithub.com\u002Fam15h\u002Ftflite_flutter_plugin\u002Ftree\u002Fmaster\u002Fexample) | tf.org & 社区 |\n| 文本检测 | CRAFT 文本检测器 ([论文](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1904.01941)) |[下载](https:\u002F\u002Fgithub.com\u002Ftulasiram58827\u002Fcraft_tflite\u002Fblob\u002Fmain\u002Fmodels\u002Fcraft_float_800.tflite?raw=true) \\| [项目仓库](https:\u002F\u002Fgithub.com\u002Ftulasiram58827\u002Fcraft_tflite\u002F) \\| [博客1-TFLite转换](https:\u002F\u002Ftulasi.dev\u002Fcraft-in-tflite) \\| [博客2-EAST vs CRAFT](https:\u002F\u002Fsayak.dev\u002Foptimizing-text-detectors\u002F) \\| [TF Hub上的模型](https:\u002F\u002Ftfhub.dev\u002Ftulasiram58827\u002Flite-model\u002Fcraft-text-detector\u002Fdr\u002F1) \\| Android（即将推出） | 社区 |\n| 文本检测 | EAST 文本检测器 ([论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F1704.03155)) |[TF Hub上的模型](https:\u002F\u002Ftfhub.dev\u002Fsayakpaul\u002Flite-model\u002Feast-text-detector\u002Fdr\u002F1) \\| [转换与推理笔记本](https:\u002F\u002Fcolab.research.google.com\u002Fgithub\u002Fsayakpaul\u002FAdventures-in-TensorFlow-Lite\u002Fblob\u002Fmaster\u002FEAST_TFLite.ipynb) | 社区 |\n\n### 语音\n| 任务 | 模型 | 应用 \\| 参考 | 来源 |\n| ------------------ |------------------------------------| ------------------------------------------------------------------------------------- | ------------ |\n| 语音识别 | DeepSpeech | [参考](https:\u002F\u002Fgithub.com\u002Fmozilla\u002FDeepSpeech\u002Ftree\u002Fmaster\u002Fnative_client\u002Fjava) | Mozilla |\n| 语音识别 | CONFORMER | [推理](https:\u002F\u002Fgithub.com\u002Fneso613\u002FASR_TFLite) [Android](https:\u002F\u002Fgithub.com\u002Fwindmaple\u002Ftflite-asr) | 社区 |\n| 语音合成 | Tacotron-2、FastSpeech2、MB-Melgan | [Android](https:\u002F\u002Fgithub.com\u002FTensorSpeech\u002FTensorflowTTS\u002Ftree\u002Fmaster\u002Fexamples\u002Fandroid) | TensorSpeech |\n| 语音合成(TTS) | Tacotron2、FastSpeech2、MelGAN、MB-MelGAN、HiFi-GAN、Parallel WaveGAN | [推理笔记本](https:\u002F\u002Fgithub.com\u002Ftulasiram58827\u002FTTS_TFLite\u002Fblob\u002Fmain\u002FEnd_to_End_TTS.ipynb) \\| [项目仓库](https:\u002F\u002Fgithub.com\u002Ftulasiram58827\u002FTTS_TFLite\u002F) | 社区 |\n\n### 推荐\n| 任务 | 模型 | 应用 \\| 参考 | 来源 |\n| ------------------ |------------------------------------| ------------------------------------------------------------------------------------- | ------------ |\n| 设备端推荐 | [双编码器](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fexamples\u002Ftree\u002Fmaster\u002Flite\u002Fexamples\u002Frecommendation\u002Fml) | [Android](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fexamples\u002Ftree\u002Fmaster\u002Flite\u002Fexamples\u002Frecommendation\u002Fandroid) \\| [iOS](https:\u002F\u002Fgithub.com\u002Fzhuzilin\u002Fon-device_recommendation_tflite) \\| [参考](https:\u002F\u002Fblog.tensorflow.org\u002F2020\u002F09\u002Fintroduction-to-tflite-on-device-recommendation.html) | tf.org & 社区 |\n\n### 游戏\n| 任务 | 模型 | 应用 \\| 参考 | 来源 |\n| ------------------ |------------------------------------| ------------------------------------------------------------------------------------- | ------------ |\n| 游戏智能体 | 强化学习 | [Flutter](https:\u002F\u002Fgithub.com\u002Fwindmaple\u002Fplanestrike-flutter) \\| [教程](https:\u002F\u002Fwindmaple.medium.com\u002F) | 社区 |\n\n\n\n## 模型库\n\n### TensorFlow Lite 模型\n这些是可以在应用和设备中实现的 TensorFlow Lite 模型：\n* [MobileNet](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodels\u002Fblob\u002Fmaster\u002Fresearch\u002Fslim\u002Fnets\u002Fmobilenet\u002FREADME.md) - 预训练的 MobileNet v2 和 v3 模型。\n* TensorFlow Lite 模型\n * [TensorFlow Lite 模型](https:\u002F\u002Fwww.tensorflow.org\u002Flite\u002Fmodels) - 包含官方的 Android 和 iOS 示例。\n * [预训练模型](https:\u002F\u002Fwww.tensorflow.org\u002Flite\u002Fguide\u002Fhosted_models) - 量化和浮点版本。\n * [TensorFlow Hub](https:\u002F\u002Ftfhub.dev\u002F) - 设置“模型格式 = TFLite”即可找到 TensorFlow Lite 模型。\n\n### TensorFlow 模型\n这些是可转换为 .tflite 格式并在应用和设备中实现的 TensorFlow 模型：\n* [TensorFlow 模型](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodels\u002Ftree\u002Fmaster\u002Fofficial) - 官方 TensorFlow 模型。\n* [TensorFlow 检测模型库](https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fmodels\u002Fblob\u002Fmaster\u002Fresearch\u002Fobject_detection\u002Fg3doc\u002Ftf2_detection_zoo.md) - 在 COCO、KITTI、AVA v2.1、iNaturalist 物种等数据集上预训练。\n\n## 思路与灵感\n* [E2E TFLite 教程](https:\u002F\u002Fgithub.com\u002Fml-gde\u002Fe2e-tflite-tutorials) - 查看此仓库，获取示例应用创意，并在你的教程项目中寻求帮助。一旦项目完成，相关的 TensorFlow Lite 模型链接、示例代码和教程将被添加到这个精彩的列表中。\n\n## ML Kit 示例\n[ML Kit](https:\u002F\u002Fdevelopers.google.com\u002Fml-kit) 是一款移动 SDK，它将 Google 的机器学习专长带给移动开发者。\n* 2019-10-01 [ML Kit 翻译演示](https:\u002F\u002Fcodelabs.developers.google.com\u002Fcodelabs\u002Fmlkit-android-translate\u002F#0) - 一个采用 Material Design 的教程，包含 [Android](https:\u002F\u002Fgithub.com\u002Fgooglecodelabs\u002Fmlkit-android\u002Ftree\u002Fmaster\u002Ftranslate)（Kotlin）示例代码 - 使用 Firebase 的 ML Kit 实时从摄像头中识别、检测语言并翻译文本。\n* 2019-03-13 [使用 ML Kit 进行计算机视觉 - Flutter 聚焦](https:\u002F\u002Fyoutu.be\u002FymyYUCrJnxU)。\n* 2019-02-09 [Flutter + MLKit：名片邮件提取器](https:\u002F\u002Fmedium.com\u002Fflutter-community\u002Fflutter-mlkit-8039ec66b6a) - 一篇带有 [Flutter](https:\u002F\u002Fgithub.com\u002FDaemonLoki\u002FBusiness-Card-Mail-Extractor) 示例代码的博客文章。\n* 2019-02-08 [从 TensorFlow 到 ML Kit：用机器学习为你的 Android 应用赋能](https:\u002F\u002Fspeakerdeck.com\u002Fjinqian\u002Ffrom-tensorflow-to-ml-kit-power-your-android-application-with-machine-learning) - 一场包含 [Android](https:\u002F\u002Fgithub.com\u002Fxebia-france\u002Fmagritte)（Kotlin）示例代码的演讲。\n* 2018-08-07 [在 Android 上使用 TensorFlow Lite 构建自定义机器学习模型](https:\u002F\u002Fmedium.com\u002Fover-engineering\u002Fbuilding-a-custom-machine-learning-model-on-android-with-tensorflow-lite-26447e53abf2)。\n* 2018-07-20 [Flutter 中的 ML Kit 和人脸检测](https:\u002F\u002Fflatteredwithflutter.com\u002Fml-kit-and-face-detection-in-flutter\u002F)。\n* 2018-07-27 [Android 上的 ML Kit 4：地标检测](https:\u002F\u002Fmedium.com\u002Fgoogle-developer-experts\u002Fexploring-firebase-mlkit-on-android-landmark-detection-part-four-5e86b8deac3a)。\n* 2018-07-28 [Android 上的 ML Kit 3：条形码扫描](https:\u002F\u002Fmedium.com\u002Fgoogle-developer-experts\u002Fexploring-firebase-mlkit-on-android-barcode-scanning-part-three-cc6f5921a108)。\n* 2018-05-31 [Android 上的 ML Kit 2：人脸检测](https:\u002F\u002Fmedium.com\u002Fgoogle-developer-experts\u002Fexploring-firebase-mlkit-on-android-face-detection-part-two-de7e307c52e0)。\n* 2018-05-22 [Android 上的 ML Kit 1：简介](https:\u002F\u002Fmedium.com\u002Fgoogle-developer-experts\u002Fexploring-firebase-mlkit-on-android-introducing-mlkit-part-one-98fcfedbeee0)。\n\n## 插件和 SDK\n* [Edge Impulse](https:\u002F\u002Fwww.edgeimpulse.com\u002F) - 由 [@EdgeImpulse](https:\u002F\u002Ftwitter.com\u002FEdgeImpulse) 创建，帮助你在云端为嵌入式设备训练 TensorFlow Lite 模型。\n* [MediaPipe](https:\u002F\u002Fgithub.com\u002Fgoogle\u002Fmediapipe) - Google AI 推出的跨平台（移动、桌面及 Edge TPU）AI 流水线。（PM [Ming Yong](https:\u002F\u002Ftwitter.com\u002Frealmgyong)) | [MediaPipe 示例](https:\u002F\u002Fmediapipe.readthedocs.io\u002Fen\u002Flatest\u002Fexamples.html)。\n* [Coral Edge TPU](https:\u002F\u002Fcoral.ai\u002F) - Google 提供的边缘硬件。[Coral Edge TPU 示例](https:\u002F\u002Fcoral.ai\u002Fexamples\u002F)。\n* [TensorFlow Lite Flutter 插件](https:\u002F\u002Fgithub.com\u002Fam15h\u002Ftflite_flutter_plugin\u002F) - 提供与 TensorFlow Lite Java API 类似的 Dart API，用于在 Flutter 应用中访问 TensorFlow Lite 解释器并执行推理。[tflite_flutter 在 pub.dev 上](https:\u002F\u002Fpub.dev\u002Fpackages\u002Ftflite_flutter)。\n\n## 有用链接\n* [Netron](https:\u002F\u002Fgithub.com\u002Flutzroeder\u002Fnetron) - 一个用于可视化模型的工具。\n* [AI 基准测试](http:\u002F\u002Fai-benchmark.com\u002Ftests.html) - 一个用于在智能手机上对计算机视觉模型进行基准测试的网站。\n* [性能测量](https:\u002F\u002Fwww.tensorflow.org\u002Flite\u002Fperformance\u002Fmeasurement) - 如何在 Android 和 iOS 上测量模型性能。\n* [ML 的 Material Design 指南](https:\u002F\u002Fmaterial.io\u002Fcollections\u002Fmachine-learning\u002Fpatterns-for-machine-learning-powered-features.html) - 如何设计基于机器学习的功能。一个很好的例子：[ML Kit 展示应用](https:\u002F\u002Fgithub.com\u002Ffirebase\u002Fmlkit-material-android)。\n* [人与 AI 指南手册](https:\u002F\u002Fpair.withgoogle.com\u002F) - 学习如何设计以人为本的 AI 产品。\n* [TensorFlow Lite 冒险之旅](https:\u002F\u002Fgithub.com\u002Fsayakpaul\u002FAdventures-in-TensorFlow-Lite) - 一个仓库，展示了非平凡的转换过程以及在 TensorFlow Lite 中的一般探索。\n* [TFProfiler](https:\u002F\u002Fgithub.com\u002Figlaweb\u002FTFProfiler) - 一款基于 Android 的应用，用于剖析 TensorFlow Lite 模型并在智能手机上测量其性能。\n* [适用于微控制器的 TensorFlow Lite](https:\u002F\u002Fwww.tensorflow.org\u002Flite\u002Fmicrocontrollers)\n* [TensorFlow Lite 示例 - Android](https:\u002F\u002Fgithub.com\u002Fdailystudio\u002Ftensorflow-lite-examples-android) - 一个仓库，重构并重写了 TensorFlow 官方网站中包含的所有 TensorFlow Lite Android 示例。\n* [TensorFlow Lite Kotlin 示例](https:\u002F\u002Fgithub.com\u002FSunitRoy2703\u002FTensorflow-lite-kotlin-samples) - 一系列用 Kotlin 编写的 TensorFlow Lite Android 示例应用，展示了不同类型的 Kotlin 实现方式，参考了 [示例应用](https:\u002F\u002Fwww.tensorflow.org\u002Flite\u002Fexamples)。\n\n\n## 学习资源\n感兴趣但不知道从哪里开始？以下是一些学习资源，无论你是初学者还是已经在该领域工作了一段时间的从业者，都能帮助你。\n\n### 博客文章\n\n* 2021年11月9日 [TensorFlow Lite 中的设备端训练](https:\u002F\u002Fblog.tensorflow.org\u002F2021\u002F11\u002Fon-device-training-in-tensorflow-lite.html)\n* 2021年9月27日 [使用 TensorFlow Lite 进行光学字符识别：一款新的示例应用](https:\u002F\u002Fblog.tensorflow.org\u002F2021\u002F09\u002Fblog.tensorflow.org202109optical-character-recognition.html)\n* 2021年6月16日 [https:\u002F\u002Fblog.tensorflow.org\u002F2021\u002F06\u002Feasier-object-detection-on-mobile-with-tf-lite.html](https:\u002F\u002Fblog.tensorflow.org\u002F2021\u002F11\u002Fon-device-training-in-tensorflow-lite.html)\n* 2020年12月29日 [YOLOv3 到 TensorFlow Lite 的转换](https:\u002F\u002Fmedium.com\u002Fanalytics-vidhya\u002Fyolov3-to-tensorflow-lite-conversion-4602cec5c239) - 作者：Nitin Tiwari。\n* 2020年4月20日 [TensorFlow Lite 的新特性](https:\u002F\u002Fblog.tensorflow.org\u002F2020\u002F04\u002Fwhats-new-in-tensorflow-lite-from-devsummit-2020.html) - 作者：Khanh LeViet。\n* 2020年4月17日 [使用 TFLite 优化风格迁移以在移动设备上运行](https:\u002F\u002Fblog.tensorflow.org\u002F2020\u002F04\u002Foptimizing-style-transfer-to-run-on-mobile-with-tflite.html) - 作者：Khanh LeViet 和 Luiz Gustavo Martins。\n* 2020年4月14日 [TensorFlow Lite 如何帮助您从原型开发到产品落地](https:\u002F\u002Fblog.tensorflow.org\u002F2020\u002F04\u002Fhow-tensorflow-lite-helps-you-from-prototype-to-product.html) - 作者：Khanh LeViet。\n* 2019年11月8日 [使用 TensorFlow 在微控制器上开始机器学习](https:\u002F\u002Fblog.particle.io\u002F2019\u002F11\u002F08\u002Fparticle-machine-learning-101\u002F) - 作者：Brandon Satrom。\n* 2019年8月5日 [TensorFlow 模型优化工具包 — float16 量化使模型大小减半](https:\u002F\u002Fblog.tensorflow.org\u002F2019\u002F08\u002Ftensorflow-model-optimization-toolkit_5.html) - 由 TensorFlow 团队撰写。\n* 2018年7月13日 [使用 Cloud TPU 在 30 分钟内训练并部署实时移动目标检测器](https:\u002F\u002Fblog.tensorflow.org\u002F2018\u002F07\u002Ftraining-and-serving-realtime-mobile-object-detector-cloud-tpus.html) - 作者：Sara Robinson、Aakanksha Chowdhery 和 Jonathan Huang。\n* 2018年6月11日 [为什么机器学习的未来是微型化的](https:\u002F\u002Fpetewarden.com\u002F2018\u002F06\u002F11\u002Fwhy-the-future-of-machine-learning-is-tiny\u002F) - 作者：Pete Warden。\n* 2018年3月30日 [在 Android 上使用 TensorFlow Lite](https:\u002F\u002Fblog.tensorflow.org\u002F2018\u002F03\u002Fusing-tensorflow-lite-on-android.html) - 作者：Laurence Moroney。\n\n### 图书\n* 2021年12月1日 [设备端 AI 和机器学习开发](https:\u002F\u002Flearning.oreilly.com\u002Flibrary\u002Fview\u002Fai-and-machine\u002F9781098101732\u002F)（抢先体验） - 作者：Laurence Moroney ([@lmoroney](https:\u002F\u002Ftwitter.com\u002Flmoroney))。\n* 2020年10月1日 [面向程序员的 AI 和机器学习](https:\u002F\u002Flearning.oreilly.com\u002Flibrary\u002Fview\u002Fai-and-machine\u002F9781492078180\u002F) - 作者：Laurence Moroney ([@lmoroney](https:\u002F\u002Ftwitter.com\u002Flmoroney))。\n* 2020年4月6日 [使用 TensorFlow Lite、ML Kit 和 Flutter 进行移动深度学习](https:\u002F\u002Fwww.packtpub.com\u002Fproduct\u002Fmobile-deep-learning-with-tensorflow-lite-ml-kit-and-flutter\u002F9781789611212)：构建可扩展的真实世界项目，以在 Android 和 iOS 上实现端到端神经网络（[GitHub](https:\u002F\u002Fgithub.com\u002FPacktPublishing\u002FMobile-Deep-Learning-Projects)） - 作者：Anubhav Singh ([@xprilion](https:\u002F\u002Fgithub.com\u002Fxprilion)) 和 Rimjhim Bhadani ([@Rimjhim28](https:\u002F\u002Fgithub.com\u002FRimjhim28))。\n* 2020年3月1日 [用于计算机视觉的 Raspberry Pi]（[完整套装](https:\u002F\u002Fwww.pyimagesearch.com\u002Fraspberry-pi-for-computer-vision) | [目录](https:\u002F\u002Fwww.pyimagesearch.com\u002F2019\u002F04\u002F05\u002Ftable-of-contents-raspberry-pi-for-computer-vision\u002F)) - 由 PyImageSearch 团队撰写：Adrian Rosebrock ([@PyImageSearch](https:\u002F\u002Ftwitter.com\u002FPyImageSearch))、David Hoffman、Asbhishek Thanki、Sayak Paul ([@RisingSayak](https:\u002F\u002Ftwitter.com\u002FRisingSayak)) 和 David Mcduffee。\n* 2019年12月1日 [TinyML](http:\u002F\u002Fshop.oreilly.com\u002Fproduct\u002F0636920254508.do) - 作者：Pete Warden ([@petewarden](https:\u002F\u002Ftwitter.com\u002Fpetewarden)) 和 Daniel Situnayake ([@dansitu](https:\u002F\u002Ftwitter.com\u002Fdansitu))。\n* 2019年10月1日 [云端、移动和边缘设备上的实用深度学习](https:\u002F\u002Fwww.practicaldeeplearning.ai\u002F) - 作者：Anirudh Koul ([@AnirudhKoul](https:\u002F\u002Ftwitter.com\u002FAnirudhKoul))、Siddha Ganju ([@SiddhaGanju](https:\u002F\u002Ftwitter.com\u002FSiddhaGanju)) 和 Meher Kasam ([@MeherKasam](https:\u002F\u002Ftwitter.com\u002FMeherKasam))。\n\n### 视频\n* 2021年10月6日 [与 Sunit Roy 一起为 TensorFlow Lite 做贡献](https:\u002F\u002Fyoutu.be\u002FsZayUoWW6nE)（Hacktoberfest 2021）\n* 2020年7月25日 [Hoi Lam 主讲的 Android 机器学习](https:\u002F\u002Fyoutu.be\u002Fm_bEh8YifnQ)（GDG Kolkata 聚会）。\n* 2020年4月1日 [轻松实现从原型到生产的设备端机器学习](https:\u002F\u002Fyoutu.be\u002FALxWJoh_BHw)（TF 开发者峰会 2020）。\n* 2020年3月11日 [TensorFlow Lite：面向移动和 IoT 设备的机器学习](https:\u002F\u002Fyoutu.be\u002F27Zx-4GOQA8)（TF 开发者峰会 2020）。\n* 2019年10月31日 [主题演讲 — TensorFlow Lite：面向移动和 IoT 设备的机器学习](https:\u002F\u002Fyoutu.be\u002FzjDGAiLqGk8)。\n* 2019年10月31日 [TensorFlow Lite：在设备上运行机器学习的解决方案](https:\u002F\u002Fyoutu.be\u002F0SpZy7iouFU)。\n* 2019年10月31日 [TensorFlow 模型优化：量化与剪枝](https:\u002F\u002Fyoutu.be\u002F3JWRVx1OKQQ)。\n* 2019年10月29日 [TensorFlow 内幕：TensorFlow Lite](https:\u002F\u002Fyoutu.be\u002FgHN0jDbJz8E)。\n* 2018年4月18日 [Android 上的 TensorFlow Lite（编码 TensorFlow）](https:\u002F\u002Fyoutu.be\u002FJnhW5tQ_7Vo)。\n\n### 播客\n* 2020年8月8日 [与 Hoi Lam 谈论机器学习](https:\u002F\u002Fanchor.fm\u002Ftalkingwithapples\u002Fepisodes\u002FTalking-Machine-Learning-with-Hoi-Lam-eiaj7v)。\n\n### MOOC 课程\n* [TensorFlow Lite 入门](https:\u002F\u002Fwww.udacity.com\u002Fcourse\u002Fintro-to-tensorflow-lite--ud190) - Udacity 课程，由 Daniel Situnayake (@dansitu)、Paige Bailey ([@DynamicWebPaige](https:\u002F\u002Ftwitter.com\u002FDynamicWebPaige)) 和 Juan Delgado 讲授。\n* [基于设备的模型与 TensorFlow Lite](https:\u002F\u002Fwww.coursera.org\u002Flearn\u002Fdevice-based-models-tensorflow) - Coursera 课程，由 Laurence Moroney ([@lmoroney](https:\u002F\u002Ftwitter.com\u002Flmoroney)) 讲授。\n* [ML 的未来是微型而光明的](https:\u002F\u002Fwww.edx.org\u002Fprofessional-certificate\u002Fharvardx-tiny-machine-learning) - 由哈佛大学与 Google 合作创建的一系列 edX 课程。讲师：Vijay Janapa Reddi、Laurence Moroney 和 Pete Warden。","# Awesome TensorFlow Lite 快速上手指南\n\n`awesome-tensorflow-lite` 并非一个可直接安装的软件包或库，而是一个精选的资源列表，汇集了适用于移动端和边缘设备的 TensorFlow Lite (TFLite) 模型、示例应用、工具及学习资源。本指南将帮助你利用该列表中的资源，快速在项目中集成 TFLite 模型。\n\n## 环境准备\n\n在开始之前，请确保你的开发环境满足以下要求：\n\n* **操作系统**: Windows, macOS, 或 Linux。\n* **Python 环境**: 推荐 Python 3.8+（用于模型转换和脚本运行）。\n* **核心依赖**:\n * TensorFlow 2.x (用于训练和模型转换)\n * TensorFlow Lite Support (可选，简化移动端开发)\n* **移动开发环境** (如需运行示例 App):\n * **Android**: Android Studio (最新版), JDK 11+\n * **iOS**: Xcode (最新版), CocoaPods\n * **Raspberry Pi**: Raspbian OS, Python 环境\n\n**安装前置依赖：**\n\n```bash\npip install tensorflow tensorflow-lite-support\n```\n\n> **提示**：国内开发者可使用清华源加速安装：\n> ```bash\n> pip install tensorflow tensorflow-lite-support -i https:\u002F\u002Fpypi.tuna.tsinghua.edu.cn\u002Fsimple\n> ```\n\n## 获取资源与示例\n\n由于这是一个资源列表，\"安装\"步骤实际上是克隆相关示例仓库或下载预训练模型。\n\n1. **浏览资源列表**：\n 访问 [awesome-tensorflow-lite GitHub 页面](https:\u002F\u002Fgithub.com\u002Fmargaretmz\u002Fawesome-tensorflow-lite) 查找你需要的任务类型（如图像分类、物体检测、文本处理等）。\n\n2. **克隆官方示例仓库** (以图像分类为例)：\n 大多数高质量示例托管在 `tensorflow\u002Fexamples` 仓库中。\n\n ```bash\n git clone https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Fexamples.git\n cd examples\u002Flite\u002Fexamples\u002Fimage_classification\n ```\n\n > **加速提示**：如果 GitHub 克隆速度慢，可使用国内镜像：\n > ```bash\n > git clone https:\u002F\u002Fgitee.com\u002Fmirrors\u002Ftensorflow-examples.git\n > ```\n\n3. **下载预训练模型**：\n 根据 README 中的表格，下载对应的 `.tflite` 模型文件。例如 MobileNetV1：\n * 链接：`https:\u002F\u002Fstorage.googleapis.com\u002Fdownload.tensorflow.org\u002Fmodels\u002Ftflite\u002Fmobilenet_v1_1.0_224_quant_and_labels.zip`\n * 解压后将 `.tflite` 文件和标签文件放入示例项目的 `assets` 目录中。\n\n## 基本使用\n\n以下展示如何在 Python 环境中加载并运行一个下载的 TFLite 模型（以图像分类为例）。这是最通用的集成方式，逻辑同样适用于 Android (Java\u002FKotlin) 和 iOS (Swift\u002FObjective-C)。\n\n### 1. 准备输入数据\n确保输入图像经过预处理（调整大小、归一化），符合模型要求。\n\n### 2. 运行推理代码\n\n```python\nimport numpy as np\nimport tensorflow as tf\n\n# 1. 加载 TFLite 模型\ninterpreter = tf.lite.Interpreter(model_path=\"mobilenet_v1_1.0_224_quant.tflite\")\ninterpreter.allocate_tensors()\n\n# 2. 获取输入输出细节\ninput_details = interpreter.get_input_details()\noutput_details = interpreter.get_output_details()\n\n# 3. 准备输入数据 (此处假设已有一个预处理好的 numpy 数组 input_data)\n# input_data 形状应为 [1, 224, 224, 3]，数据类型需与 input_details 匹配\ninput_data = np.random.rand(1, 224, 224, 3).astype(np.float32) \n\n# 4. 设置输入张量\ninterpreter.set_tensor(input_details[0]['index'], input_data)\n\n# 5. 运行推理\ninterpreter.invoke()\n\n# 6. 获取输出结果\noutput_data = interpreter.get_tensor(output_details[0]['index'])\n\n# 7. 后处理 (例如获取概率最高的类别索引)\npredicted_class = np.argmax(output_data[0])\nprint(f\"Predicted class index: {predicted_class}\")\n```\n\n### 移动端集成简述\n* **Android**: 将 `.tflite` 文件放入 `app\u002Fsrc\u002Fmain\u002Fassets`，参考 `examples\u002Flite\u002Fexamples\u002Fimage_classification\u002Fandroid` 中的 Java\u002FKotlin 代码调用 `Interpreter` API。\n* **iOS**: 将 `.tflite` 文件拖入 Xcode 项目，参考 `examples\u002Flite\u002Fexamples\u002Fimage_classification\u002Fios` 中的 Swift 代码使用 `TensorFlowLiteSwift` 库。\n\n通过查阅 `awesome-tensorflow-lite` 列表中的具体条目，你可以找到针对特定任务（如姿态估计、风格迁移）的完整工程代码和更详细的教程链接。","一家初创团队正致力于开发一款面向户外爱好者的智能眼镜，需要在设备端实时运行姿态估计和图像分类模型，且必须保证低延迟以节省电量。\n\n### 没有 awesome-tensorflow-lite 时\n- **资源分散难查找**：开发者需在 GitHub、博客和技术论坛间反复搜索，难以确认哪些预训练模型真正支持移动端部署。\n- **示例代码缺失**：找到模型后，往往缺乏配套的 Android\u002FiOS 示例代码，导致从模型转换到集成的每一步都要从头摸索。\n- **技术选型盲目**：不清楚社区是否有现成的工具（如 Model Maker 或 Hexagon Delegate）来优化推理速度，只能手动尝试各种参数，效率极低。\n- **学习曲线陡峭**：缺乏系统化的教程和视频资源，团队成员在理解 TFLite 新特性（如端侧训练）时耗费大量时间试错。\n\n### 使用 awesome-tensorflow-lite 后\n- **一站式资源聚合**：直接查阅分类清晰的模型列表，快速锁定适合户外场景的轻量级姿态估计模型及其官方样例。\n- **开箱即用的示例**：每个模型都附带完整的示例应用代码，团队可直接复用图像分类和检测模块，将集成周期从数周缩短至几天。\n- **性能优化有据可依**：通过\"Plugins and SDKs\"板块迅速定位 Hexagon Delegate 等加速方案，显著提升了模型在边缘设备上的推理帧率。\n- **系统化学习路径**：利用精选的博客、视频和 MOOC 资源，团队快速掌握了端侧个性化训练技术，实现了用户习惯的本地自适应调整。\n\nawesome-tensorflow-lite 通过将分散的社区智慧结构化，让移动端 AI 开发从“大海捞针”变成了“按图索骥”，极大降低了边缘计算落地的门槛。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fmargaretmz_awesome-tensorflow-lite_d02815ab.png","margaretmz","Margaret Maynard-Reid","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002Fmargaretmz_de3965a0.png","AI & Cloud GDE | AI, Art & Design",null,"Seattle, US","https:\u002F\u002Fmargaretmz.me","https:\u002F\u002Fgithub.com\u002Fmargaretmz",1372,190,"2026-03-23T07:54:49","Apache-2.0",1,"Android, iOS, Linux (Raspberry Pi), 未说明 (Windows\u002FmacOS 桌面端支持取决于具体示例实现)","非必需。主要面向移动端和边缘设备 CPU\u002FNPU\u002FDSP 推理。支持通过 Hexagon Delegate 在 Qualcomm DSP 上加速，未提及特定桌面 GPU 或 CUDA 版本需求。","未说明 (取决于具体模型和设备，通常为移动端低内存环境)",{"notes":89,"python":90,"dependencies":91},"该仓库是一个资源列表而非单一可执行软件。它汇集了适用于移动设备（Android\u002FiOS）和边缘设备（如树莓派）的 TensorFlow Lite 模型、示例应用及工具。运行环境高度依赖于具体的示例项目（如需要 Android Studio 开发安卓应用，或 Xcode 开发 iOS 应用）。部分功能支持利用 Qualcomm Hexagon DSP 进行硬件加速。","未说明 (核心库为 C++，但模型训练和转换示例通常使用 Python 3.x)",[92,93,94,95,96],"TensorFlow Lite","TensorFlow (用于模型训练和转换)","MediaPipe (部分模型依赖)","Android Support Library (可选)","Model Maker (可选)",[52,15,14],[99,100,101,102,103,104,105,106,107,108,109,110,111,112,113,114,115,116,117,118],"awesome-list","tflite","tensorflow-lite","tflite-models","tensorflow","tensorflow-models","tfhub","tensorflow-keras","keras-tutorials","sample-app","model-zoo","mediapipe","mlkit","mobile","android","ios","flutter","deep-learning","computer-vision","awesome","2026-03-27T02:49:30.150509","2026-04-15T01:04:41.254205",[],[]]