[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"similar-JaidedAI--EasyOCR":3,"tool-JaidedAI--EasyOCR":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",140436,2,"2026-04-05T23:32:43",[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。",107662,"2026-04-03T11:11:01",[14,15,13],{"id":45,"name":46,"github_repo":47,"description_zh":48,"stars":49,"difficulty_score":10,"last_commit_at":50,"category_tags":51,"status":17},4292,"Deep-Live-Cam","hacksider\u002FDeep-Live-Cam","Deep-Live-Cam 是一款专注于实时换脸与视频生成的开源工具，用户仅需一张静态照片，即可通过“一键操作”实现摄像头画面的即时变脸或制作深度伪造视频。它有效解决了传统换脸技术流程繁琐、对硬件配置要求极高以及难以实时预览的痛点，让高质量的数字内容创作变得触手可及。\n\n这款工具不仅适合开发者和技术研究人员探索算法边界，更因其极简的操作逻辑（仅需三步：选脸、选摄像头、启动），广泛适用于普通用户、内容创作者、设计师及直播主播。无论是为了动画角色定制、服装展示模特替换，还是制作趣味短视频和直播互动，Deep-Live-Cam 都能提供流畅的支持。\n\n其核心技术亮点在于强大的实时处理能力，支持口型遮罩（Mouth Mask）以保留使用者原始的嘴部动作，确保表情自然精准；同时具备“人脸映射”功能，可同时对画面中的多个主体应用不同面孔。此外，项目内置了严格的内容安全过滤机制，自动拦截涉及裸露、暴力等不当素材，并倡导用户在获得授权及明确标注的前提下合规使用，体现了技术发展与伦理责任的平衡。",88924,"2026-04-06T03:28:53",[14,15,13,52],"视频",{"id":54,"name":55,"github_repo":56,"description_zh":57,"stars":58,"difficulty_score":32,"last_commit_at":59,"category_tags":60,"status":17},3704,"NextChat","ChatGPTNextWeb\u002FNextChat","NextChat 是一款轻量且极速的 AI 助手，旨在为用户提供流畅、跨平台的大模型交互体验。它完美解决了用户在多设备间切换时难以保持对话连续性，以及面对众多 AI 模型不知如何统一管理的痛点。无论是日常办公、学习辅助还是创意激发，NextChat 都能让用户随时随地通过网页、iOS、Android、Windows、MacOS 或 Linux 端无缝接入智能服务。\n\n这款工具非常适合普通用户、学生、职场人士以及需要私有化部署的企业团队使用。对于开发者而言，它也提供了便捷的自托管方案，支持一键部署到 Vercel 或 Zeabur 等平台。\n\nNextChat 的核心亮点在于其广泛的模型兼容性，原生支持 Claude、DeepSeek、GPT-4 及 Gemini Pro 等主流大模型，让用户在一个界面即可自由切换不同 AI 能力。此外，它还率先支持 MCP（Model Context Protocol）协议，增强了上下文处理能力。针对企业用户，NextChat 提供专业版解决方案，具备品牌定制、细粒度权限控制、内部知识库整合及安全审计等功能，满足公司对数据隐私和个性化管理的高标准要求。",87618,"2026-04-05T07:20:52",[14,35],{"id":62,"github_repo":63,"name":64,"description_en":65,"description_zh":66,"ai_summary_zh":66,"readme_en":67,"readme_zh":68,"quickstart_zh":69,"use_case_zh":70,"hero_image_url":71,"owner_login":72,"owner_name":73,"owner_avatar_url":74,"owner_bio":75,"owner_company":76,"owner_location":76,"owner_email":76,"owner_twitter":72,"owner_website":77,"owner_url":78,"languages":79,"stars":112,"forks":113,"last_commit_at":114,"license":115,"difficulty_score":32,"env_os":116,"env_gpu":117,"env_ram":118,"env_deps":119,"category_tags":126,"github_topics":128,"view_count":32,"oss_zip_url":76,"oss_zip_packed_at":76,"status":17,"created_at":144,"updated_at":145,"faqs":146,"releases":175},4264,"JaidedAI\u002FEasyOCR","EasyOCR","Ready-to-use OCR with 80+ supported languages and all popular writing scripts including Latin, Chinese, Arabic, Devanagari, Cyrillic and etc.","EasyOCR 是一款开箱即用的光学字符识别（OCR）工具，旨在帮助开发者轻松从图片中提取文字。它支持全球 80 多种语言及主流书写体系，涵盖拉丁文、中文、阿拉伯文、天城文和西里尔文等，能有效解决多语言混合场景下的文字识别难题。\n\n无论是需要处理国际化文档的软件开发人员，还是从事数据标注的研究者，亦或是希望快速集成 OCR 功能的设计师，都能从中受益。只需几行 Python 代码，用户即可加载模型并识别本地图片、内存图像甚至网络链接中的文字内容。\n\n其技术亮点在于极高的易用性与灵活性：首次加载模型后无需重复初始化，支持一次性传入多种兼容语言进行混合识别，并能返回包含坐标框、文本内容及置信度的详细结果，也可简化为纯文本列表输出。此外，项目自动管理模型权重下载，同时提供 Docker 部署方案与 Hugging Face 在线演示，即便在没有 GPU 的环境下也能运行。作为基于 Apache 2.0 协议开源的项目，EasyOCR 让高质量的文字识别技术变得触手可及。","# EasyOCR\n\n[![PyPI Status](https:\u002F\u002Fbadge.fury.io\u002Fpy\u002Feasyocr.svg)](https:\u002F\u002Fbadge.fury.io\u002Fpy\u002Feasyocr)\n[![license](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FLicense-Apache%202.0-blue.svg)](https:\u002F\u002Fgithub.com\u002FJaidedAI\u002FEasyOCR\u002Fblob\u002Fmaster\u002FLICENSE)\n[![Tweet](https:\u002F\u002Fimg.shields.io\u002Ftwitter\u002Furl\u002Fhttps\u002Fgithub.com\u002FJaidedAI\u002FEasyOCR.svg?style=social)](https:\u002F\u002Ftwitter.com\u002Fintent\u002Ftweet?text=Check%20out%20this%20awesome%20library:%20EasyOCR%20https:\u002F\u002Fgithub.com\u002FJaidedAI\u002FEasyOCR)\n[![Twitter](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Ftwitter-@JaidedAI-blue.svg?style=flat)](https:\u002F\u002Ftwitter.com\u002FJaidedAI)\n\nReady-to-use OCR with 80+ [supported languages](https:\u002F\u002Fwww.jaided.ai\u002Feasyocr) and all popular writing scripts including: Latin, Chinese, Arabic, Devanagari, Cyrillic, etc.\n\n[Try Demo on our website](https:\u002F\u002Fwww.jaided.ai\u002Feasyocr)\n\nIntegrated into [Huggingface Spaces 🤗](https:\u002F\u002Fhuggingface.co\u002Fspaces) using [Gradio](https:\u002F\u002Fgithub.com\u002Fgradio-app\u002Fgradio). Try out the Web Demo: [![Hugging Face Spaces](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F%F0%9F%A4%97%20Hugging%20Face-Spaces-blue)](https:\u002F\u002Fhuggingface.co\u002Fspaces\u002Ftomofi\u002FEasyOCR)\n\n\n## What's new\n- 24 September 2024 - Version 1.7.2\n    - Fix several compatibilities\n\n- [Read all release notes](https:\u002F\u002Fgithub.com\u002FJaidedAI\u002FEasyOCR\u002Fblob\u002Fmaster\u002Freleasenotes.md)\n\n## What's coming next\n- Handwritten text support\n\n## Examples\n\n![example](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FJaidedAI_EasyOCR_readme_392e2f04db00.png)\n\n![example2](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FJaidedAI_EasyOCR_readme_3a1f05b2b4c9.png)\n\n![example3](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FJaidedAI_EasyOCR_readme_e4c43f1e5e08.png)\n\n\n## Installation\n\nInstall using `pip`\n\nFor the latest stable release:\n\n``` bash\npip install easyocr\n```\n\nFor the latest development release:\n\n``` bash\npip install git+https:\u002F\u002Fgithub.com\u002FJaidedAI\u002FEasyOCR.git\n```\n\nNote 1: For Windows, please install torch and torchvision first by following the official instructions here https:\u002F\u002Fpytorch.org. On the pytorch website, be sure to select the right CUDA version you have. If you intend to run on CPU mode only, select `CUDA = None`.\n\nNote 2: We also provide a Dockerfile [here](https:\u002F\u002Fgithub.com\u002FJaidedAI\u002FEasyOCR\u002Fblob\u002Fmaster\u002FDockerfile).\n\n## Usage\n\n``` python\nimport easyocr\nreader = easyocr.Reader(['ch_sim','en']) # this needs to run only once to load the model into memory\nresult = reader.readtext('chinese.jpg')\n```\n\nThe output will be in a list format, each item represents a bounding box, the text detected and confident level, respectively.\n\n``` bash\n[([[189, 75], [469, 75], [469, 165], [189, 165]], '愚园路', 0.3754989504814148),\n ([[86, 80], [134, 80], [134, 128], [86, 128]], '西', 0.40452659130096436),\n ([[517, 81], [565, 81], [565, 123], [517, 123]], '东', 0.9989598989486694),\n ([[78, 126], [136, 126], [136, 156], [78, 156]], '315', 0.8125889301300049),\n ([[514, 126], [574, 126], [574, 156], [514, 156]], '309', 0.4971577227115631),\n ([[226, 170], [414, 170], [414, 220], [226, 220]], 'Yuyuan Rd.', 0.8261902332305908),\n ([[79, 173], [125, 173], [125, 213], [79, 213]], 'W', 0.9848111271858215),\n ([[529, 173], [569, 173], [569, 213], [529, 213]], 'E', 0.8405593633651733)]\n```\nNote 1: `['ch_sim','en']` is the list of languages you want to read. You can pass\nseveral languages at once but not all languages can be used together.\nEnglish is compatible with every language and languages that share common characters are usually compatible with each other.\n\nNote 2: Instead of the filepath `chinese.jpg`, you can also pass an OpenCV image object (numpy array) or an image file as bytes. A URL to a raw image is also acceptable.\n\nNote 3: The line `reader = easyocr.Reader(['ch_sim','en'])` is for loading a model into memory. It takes some time but it needs to be run only once.\n\nYou can also set `detail=0` for simpler output.\n\n``` python\nreader.readtext('chinese.jpg', detail = 0)\n```\nResult:\n``` bash\n['愚园路', '西', '东', '315', '309', 'Yuyuan Rd.', 'W', 'E']\n```\n\nModel weights for the chosen language will be automatically downloaded or you can\ndownload them manually from the [model hub](https:\u002F\u002Fwww.jaided.ai\u002Feasyocr\u002Fmodelhub) and put them in the '~\u002F.EasyOCR\u002Fmodel' folder\n\nIn case you do not have a GPU, or your GPU has low memory, you can run the model in CPU-only mode by adding `gpu=False`.\n\n``` python\nreader = easyocr.Reader(['ch_sim','en'], gpu=False)\n```\n\nFor more information, read the [tutorial](https:\u002F\u002Fwww.jaided.ai\u002Feasyocr\u002Ftutorial) and [API Documentation](https:\u002F\u002Fwww.jaided.ai\u002Feasyocr\u002Fdocumentation).\n\n#### Run on command line\n\n```shell\n$ easyocr -l ch_sim en -f chinese.jpg --detail=1 --gpu=True\n```\n\n## Train\u002Fuse your own model\n\nFor recognition model, [Read here](https:\u002F\u002Fgithub.com\u002FJaidedAI\u002FEasyOCR\u002Fblob\u002Fmaster\u002Fcustom_model.md).\n\nFor detection model (CRAFT), [Read here](https:\u002F\u002Fgithub.com\u002FJaidedAI\u002FEasyOCR\u002Fblob\u002Fmaster\u002Ftrainer\u002Fcraft\u002FREADME.md).\n\n## Implementation Roadmap\n\n- Handwritten support\n- Restructure code to support swappable detection and recognition algorithms\nThe api should be as easy as\n``` python\nreader = easyocr.Reader(['en'], detection='DB', recognition = 'Transformer')\n```\nThe idea is to be able to plug in any state-of-the-art model into EasyOCR. There are a lot of geniuses trying to make better detection\u002Frecognition models, but we are not trying to be geniuses here. We just want to make their works quickly accessible to the public ... for free. (well, we believe most geniuses want their work to create a positive impact as fast\u002Fbig as possible) The pipeline should be something like the below diagram. Grey slots are placeholders for changeable light blue modules.\n\n![plan](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FJaidedAI_EasyOCR_readme_13a5d0072cf6.jpeg)\n\n## Acknowledgement and References\n\nThis project is based on research and code from several papers and open-source repositories.\n\nAll deep learning execution is based on [Pytorch](https:\u002F\u002Fpytorch.org). :heart:\n\nDetection execution uses the CRAFT algorithm from this [official repository](https:\u002F\u002Fgithub.com\u002Fclovaai\u002FCRAFT-pytorch) and their [paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F1904.01941) (Thanks @YoungminBaek from [@clovaai](https:\u002F\u002Fgithub.com\u002Fclovaai)). We also use their pretrained model. Training script is provided by [@gmuffiness](https:\u002F\u002Fgithub.com\u002Fgmuffiness).\n\nThe recognition model is a CRNN ([paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F1507.05717)). It is composed of 3 main components: feature extraction (we are currently using [Resnet](https:\u002F\u002Farxiv.org\u002Fabs\u002F1512.03385)) and VGG, sequence labeling ([LSTM](https:\u002F\u002Fwww.bioinf.jku.at\u002Fpublications\u002Folder\u002F2604.pdf)) and decoding ([CTC](https:\u002F\u002Fwww.cs.toronto.edu\u002F~graves\u002Ficml_2006.pdf)). The training pipeline for recognition execution is a modified version of the [deep-text-recognition-benchmark](https:\u002F\u002Fgithub.com\u002Fclovaai\u002Fdeep-text-recognition-benchmark) framework. (Thanks [@ku21fan](https:\u002F\u002Fgithub.com\u002Fku21fan) from [@clovaai](https:\u002F\u002Fgithub.com\u002Fclovaai)) This repository is a gem that deserves more recognition.\n\nBeam search code is based on this [repository](https:\u002F\u002Fgithub.com\u002Fgithubharald\u002FCTCDecoder) and his [blog](https:\u002F\u002Ftowardsdatascience.com\u002Fbeam-search-decoding-in-ctc-trained-neural-networks-5a889a3d85a7). (Thanks [@githubharald](https:\u002F\u002Fgithub.com\u002Fgithubharald))\n\nData synthesis is based on [TextRecognitionDataGenerator](https:\u002F\u002Fgithub.com\u002FBelval\u002FTextRecognitionDataGenerator). (Thanks [@Belval](https:\u002F\u002Fgithub.com\u002FBelval))\n\nAnd a good read about CTC from distill.pub [here](https:\u002F\u002Fdistill.pub\u002F2017\u002Fctc\u002F).\n\n## Want To Contribute?\n\nLet's advance humanity together by making AI available to everyone!\n\n3 ways to contribute:\n\n**Coder:** Please send a PR for small bugs\u002Fimprovements. For bigger ones, discuss with us by opening an issue first. There is a list of possible bug\u002Fimprovement issues tagged with ['PR WELCOME'](https:\u002F\u002Fgithub.com\u002FJaidedAI\u002FEasyOCR\u002Fissues?q=is%3Aissue+is%3Aopen+label%3A%22PR+WELCOME%22).\n\n**User:** Tell us how EasyOCR benefits you\u002Fyour organization to encourage further development. Also post failure cases in [Issue  Section](https:\u002F\u002Fgithub.com\u002FJaidedAI\u002FEasyOCR\u002Fissues) to help improve future models.\n\n**Tech leader\u002FGuru:** If you found this library useful, please spread the word! (See [Yann Lecun's post](https:\u002F\u002Fwww.facebook.com\u002Fyann.lecun\u002Fposts\u002F10157018122787143) about EasyOCR)\n\n## Guideline for new language request\n\nTo request a new language, we need you to send a PR with the 2 following files:\n\n1. In folder [easyocr\u002Fcharacter](https:\u002F\u002Fgithub.com\u002FJaidedAI\u002FEasyOCR\u002Ftree\u002Fmaster\u002Feasyocr\u002Fcharacter),\nwe need 'yourlanguagecode_char.txt' that contains list of all characters. Please see format examples from other files in that folder.\n2. In folder [easyocr\u002Fdict](https:\u002F\u002Fgithub.com\u002FJaidedAI\u002FEasyOCR\u002Ftree\u002Fmaster\u002Feasyocr\u002Fdict),\nwe need 'yourlanguagecode.txt' that contains list of words in your language.\nOn average, we have ~30000 words per language with more than 50000 words for more popular ones.\nMore is better in this file.\n\nIf your language has unique elements (such as 1. Arabic: characters change form when attached to each other + write from right to left 2. Thai: Some characters need to be above the line and some below), please educate us to the best of your ability and\u002For give useful links. It is important to take care of the detail to achieve a system that really works.\n\nLastly, please understand that our priority will have to go to popular languages or sets of languages that share large portions of their characters with each other (also tell us if this is the case for your language). It takes us at least a week to develop a new model, so you may have to wait a while for the new model to be released.\n\nSee [List of languages in development](https:\u002F\u002Fgithub.com\u002FJaidedAI\u002FEasyOCR\u002Fissues\u002F91)\n\n## Github Issues\n\nDue to limited resources, an issue older than 6 months will be automatically closed. Please open an issue again if it is critical.\n\n## Business Inquiries\n\nFor Enterprise Support, [Jaided AI](https:\u002F\u002Fwww.jaided.ai\u002F) offers full service for custom OCR\u002FAI systems from implementation, training\u002Ffinetuning and deployment. Click [here](https:\u002F\u002Fwww.jaided.ai\u002Fcontactus?ref=github) to contact us.\n","# EasyOCR\n\n[![PyPI 状态](https:\u002F\u002Fbadge.fury.io\u002Fpy\u002Feasyocr.svg)](https:\u002F\u002Fbadge.fury.io\u002Fpy\u002Feasyocr)\n[![许可证](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FLicense-Apache%202.0-blue.svg)](https:\u002F\u002Fgithub.com\u002FJaidedAI\u002FEasyOCR\u002Fblob\u002Fmaster\u002FLICENSE)\n[![推文](https:\u002F\u002Fimg.shields.io\u002Ftwitter\u002Furl\u002Fhttps\u002Fgithub.com\u002FJaidedAI\u002FEasyOCR.svg?style=social)](https:\u002F\u002Ftwitter.com\u002Fintent\u002Ftweet?text=Check%20out%20this%20awesome%20library:%20EasyOCR%20https:\u002F\u002Fgithub.com\u002FJaidedAI\u002FEasyOCR)\n[![Twitter](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Ftwitter-@JaidedAI-blue.svg?style=flat)](https:\u002F\u002Ftwitter.com\u002FJaidedAI)\n\n开箱即用的 OCR 工具，支持 80 多种语言及所有主流书写系统，包括拉丁字母、中文、阿拉伯文、天城文、西里尔字母等。\n\n[在我们的网站上试用演示](https:\u002F\u002Fwww.jaided.ai\u002Feasyocr)\n\n集成到 [Huggingface Spaces 🤗](https:\u002F\u002Fhuggingface.co\u002Fspaces) 中，使用 [Gradio](https:\u002F\u002Fgithub.com\u002Fgradio-app\u002Fgradio)。体验在线演示：[![Hugging Face Spaces](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F%F0%9F%A4%97%20Hugging%20Face-Spaces-blue)](https:\u002F\u002Fhuggingface.co\u002Fspaces\u002Ftomofi\u002FEasyOCR)\n\n\n## 最新动态\n- 2024年9月24日 - 版本 1.7.2\n    - 修复了若干兼容性问题\n\n- [查看全部发布说明](https:\u002F\u002Fgithub.com\u002FJaidedAI\u002FEasyOCR\u002Fblob\u002Fmaster\u002Freleasenotes.md)\n\n## 下一步计划\n- 手写文本支持\n\n## 示例\n\n![example](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FJaidedAI_EasyOCR_readme_392e2f04db00.png)\n\n![example2](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FJaidedAI_EasyOCR_readme_3a1f05b2b4c9.png)\n\n![example3](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FJaidedAI_EasyOCR_readme_e4c43f1e5e08.png)\n\n\n## 安装\n\n使用 `pip` 进行安装\n\n对于最新稳定版：\n\n``` bash\npip install easyocr\n```\n\n对于最新开发版：\n\n``` bash\npip install git+https:\u002F\u002Fgithub.com\u002FJaidedAI\u002FEasyOCR.git\n```\n\n注意 1：对于 Windows 用户，请先按照官方指南 https:\u002F\u002Fpytorch.org 安装 PyTorch 和 torchvision。在 PyTorch 官网中，请确保选择与您显卡匹配的 CUDA 版本。如果您仅打算在 CPU 上运行，请选择 `CUDA = None`。\n\n注意 2：我们还提供了一个 Dockerfile [在这里](https:\u002F\u002Fgithub.com\u002FJaidedAI\u002FEasyOCR\u002Fblob\u002Fmaster\u002FDockerfile)。\n\n## 使用方法\n\n``` python\nimport easyocr\nreader = easyocr.Reader(['ch_sim','en']) # 此步骤只需执行一次，用于将模型加载到内存中\nresult = reader.readtext('chinese.jpg')\n```\n\n输出将以列表形式呈现，每个元素分别包含检测到的文本框坐标、识别出的文本内容以及置信度。\n\n``` bash\n[([[189, 75], [469, 75], [469, 165], [189, 165]], '愚园路', 0.3754989504814148),\n ([[86, 80], [134, 80], [134, 128], [86, 128]], '西', 0.40452659130096436),\n ([[517, 81], [565, 81], [565, 123], [517, 123]], '东', 0.9989598989486694),\n ([[78, 126], [136, 126], [136, 156], [78, 156]], '315', 0.8125889301300049),\n ([[514, 126], [574, 126], [574, 156], [514, 156]], '309', 0.4971577227115631),\n ([[226, 170], [414, 170], [414, 220], [226, 220]], 'Yuyuan Rd.', 0.8261902332305908),\n ([[79, 173], [125, 173], [125, 213], [79, 213]], 'W', 0.9848111271858215),\n ([[529, 173], [569, 173], [569, 213], [529, 213]], 'E', 0.8405593633651733)]\n```\n注意 1：`['ch_sim','en']` 是您希望识别的语言列表。您可以同时指定多种语言，但并非所有语言都能组合使用。英语可以与任何语言搭配使用，而具有共同字符的语言通常也能相互兼容。\n\n注意 2：除了文件路径 `'chinese.jpg'` 外，您还可以传入 OpenCV 图像对象（numpy 数组）或以字节形式表示的图像文件。直接指向原始图像的 URL 也是可行的。\n\n注意 3：语句 `reader = easyocr.Reader(['ch_sim','en'])` 用于将模型加载到内存中。这需要一些时间，但只需执行一次即可。\n\n您也可以通过设置 `detail=0` 来获得更简洁的输出。\n\n``` python\nreader.readtext('chinese.jpg', detail = 0)\n```\n结果：\n``` bash\n['愚园路', '西', '东', '315', '309', 'Yuyuan Rd.', 'W', 'E']\n```\n\n所选语言的模型权重将会自动下载；您也可以从 [模型中心](https:\u002F\u002Fwww.jaided.ai\u002Feasyocr\u002Fmodelhub) 手动下载，并将其放置在 `~\u002F.EasyOCR\u002Fmodel` 文件夹中。\n\n如果您的设备没有 GPU，或者 GPU 内存不足，可以通过添加 `gpu=False` 参数以纯 CPU 模式运行模型。\n\n``` python\nreader = easyocr.Reader(['ch_sim','en'], gpu=False)\n```\n\n欲了解更多信息，请参阅 [教程](https:\u002F\u002Fwww.jaided.ai\u002Feasyocr\u002Ftutorial) 和 [API 文档](https:\u002F\u002Fwww.jaided.ai\u002Feasyocr\u002Fdocumentation)。\n\n#### 命令行运行\n\n```shell\n$ easyocr -l ch_sim en -f chinese.jpg --detail=1 --gpu=True\n```\n\n## 训练\u002F使用自定义模型\n\n关于识别模型，请参阅 [此处](https:\u002F\u002Fgithub.com\u002FJaidedAI\u002FEasyOCR\u002Fblob\u002Fmaster\u002Fcustom_model.md)。\n\n关于检测模型（CRAFT），请参阅 [此处](https:\u002F\u002Fgithub.com\u002FJaidedAI\u002FEasyOCR\u002Fblob\u002Fmaster\u002Ftrainer\u002Fcraft\u002FREADME.md)。\n\n## 实施路线图\n\n- 手写文本支持\n- 重构代码以支持可替换的检测和识别算法\nAPI 应该简单易用，例如：\n``` python\nreader = easyocr.Reader(['en'], detection='DB', recognition = 'Transformer')\n```\n我们的目标是能够将任何最先进的模型无缝集成到 EasyOCR 中。许多天才正在努力开发更好的检测和识别模型，但我们并不追求成为这些领域的专家。我们只想让这些优秀成果快速、免费地惠及大众……（毕竟，大多数天才都希望自己的工作能尽快、尽可能广泛地产生积极影响）。整个流程应类似于下图所示，灰色部分为可更换的浅蓝色模块。\n\n![plan](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FJaidedAI_EasyOCR_readme_13a5d0072cf6.jpeg)\n\n## 致谢与参考文献\n\n本项目基于多篇论文及开源仓库中的研究和代码。\n\n所有深度学习的执行均基于 [Pytorch](https:\u002F\u002Fpytorch.org)。:heart:\n\n检测部分采用了来自此 [官方仓库](https:\u002F\u002Fgithub.com\u002Fclovaai\u002FCRAFT-pytorch) 的 CRAFT 算法及其 [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F1904.01941)（感谢来自 [@clovaai](https:\u002F\u002Fgithub.com\u002Fclovaai) 的 @YoungminBaek）。我们还使用了他们提供的预训练模型。训练脚本由 [@gmuffiness](https:\u002F\u002Fgithub.com\u002Fgmuffiness) 提供。\n\n识别模型为 CRNN（[论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F1507.05717)），由三个主要组件构成：特征提取（目前我们使用 [Resnet](https:\u002F\u002Farxiv.org\u002Fabs\u002F1512.03385) 和 VGG）、序列标注（[LSTM](https:\u002F\u002Fwww.bioinf.jku.at\u002Fpublications\u002Folder\u002F2604.pdf)）以及解码（[CTC](https:\u002F\u002Fwww.cs.toronto.edu\u002F~graves\u002Ficml_2006.pdf)）。识别部分的训练流程是对 [deep-text-recognition-benchmark](https:\u002F\u002Fgithub.com\u002Fclovaai\u002Fdeep-text-recognition-benchmark) 框架的修改版本。（感谢来自 [@clovaai](https:\u002F\u002Fgithub.com\u002Fclovaai) 的 [@ku21fan]）该仓库堪称瑰宝，值得更多关注。\n\nBeam search 的代码基于此 [仓库](https:\u002F\u002Fgithub.com\u002Fgithubharald\u002FCTCDecoder) 及其 [博客](https:\u002F\u002Ftowardsdatascience.com\u002Fbeam-search-decoding-in-ctc-trained-neural-networks-5a889a3d85a7)。（感谢 [@githubharald](https:\u002F\u002Fgithub.com\u002Fgithubharald)）\n\n数据合成基于 [TextRecognitionDataGenerator](https:\u002F\u002Fgithub.com\u002FBelval\u002FTextRecognitionDataGenerator)。（感谢 [@Belval](https:\u002F\u002Fgithub.com\u002FBelval)）\n\n此外，distill.pub 上有一篇关于 CTC 的优秀文章，可在此查阅：[链接](https:\u002F\u002Fdistill.pub\u002F2017\u002Fctc\u002F)。\n\n## 想要贡献吗？\n\n让我们携手推动人工智能普惠化，共同造福人类！\n\n有三种方式可以参与贡献：\n\n**开发者：** 请针对小问题或改进提交 Pull Request。若涉及较大改动，请先通过新建 Issue 与我们讨论。我们已将一些可能的 Bug 或改进问题标记为 ['PR WELCOME'](https:\u002F\u002Fgithub.com\u002FJaidedAI\u002FEasyOCR\u002Fissues?q=is%3Aissue+is%3Aopen+label%3A%22PR+WELCOME%22)，欢迎随时提出。\n\n**用户：** 请告诉我们 EasyOCR 如何帮助了您或您的组织，以激励我们持续开发。同时，也欢迎您在 [Issue 栏](https:\u002F\u002Fgithub.com\u002FJaidedAI\u002FEasyOCR\u002Fissues) 中分享遇到的失败案例，这将有助于我们优化未来的模型。\n\n**技术领袖\u002F专家：** 如果您认为本库很有价值，请帮忙宣传！（参见 Yann LeCun 关于 EasyOCR 的 [帖子](https:\u002F\u002Fwww.facebook.com\u002Fyann.lecun\u002Fposts\u002F10157018122787143)）\n\n## 新语言请求指南\n\n如需申请支持新语言，您需要提交一个包含以下两个文件的 PR：\n\n1. 在 [easyocr\u002Fcharacter](https:\u002F\u002Fgithub.com\u002FJaidedAI\u002FEasyOCR\u002Ftree\u002Fmaster\u002Feasyocr\u002Fcharacter) 文件夹中，我们需要一个名为 'yourlanguagecode_char.txt' 的文件，其中列出该语言的所有字符。请参考该文件夹内其他文件的格式示例。\n   \n2. 在 [easyocr\u002Fdict](https:\u002F\u002Fgithub.com\u002FJaidedAI\u002FEasyOCR\u002Ftree\u002Fmaster\u002Feasyocr\u002Fdict) 文件夹中，我们需要一个名为 'yourlanguagecode.txt' 的文件，其中包含该语言的所有词汇。通常每种语言约有 30,000 个词汇，较为流行的语言则超过 50,000 个。文件中词汇越多越好。\n\n如果您的语言具有特殊性（例如：1. 阿拉伯语：字母连接时会改变形状，且从右向左书写；2. 泰语：部分字母需置于基线上方，部分则在下方），请尽可能详细地向我们说明，或提供相关参考资料。只有充分考虑这些细节，才能构建出真正可用的系统。\n\n最后，请理解我们的优先级将倾向于那些使用广泛的语言，或是与其他语言共享大量字符的语言组合（请告知我们您的语言是否属于此类）。开发一个新的语言模型至少需要一周时间，因此您可能需要耐心等待新模型的发布。\n\n更多正在开发的语言列表，请参阅 [此处](https:\u002F\u002Fgithub.com\u002FJaidedAI\u002FEasyOCR\u002Fissues\u002F91)。\n\n## GitHub Issues\n\n由于资源有限，超过 6 个月未更新的 Issue 将被自动关闭。若问题仍至关重要，请重新打开新的 Issue。\n\n## 商务咨询\n\n对于企业级支持，[Jaided AI](https:\u002F\u002Fwww.jaided.ai\u002F) 提供 OCR\u002FAI 系统的全方位定制服务，涵盖实施、训练\u002F微调及部署等环节。请点击 [这里](https:\u002F\u002Fwww.jaided.ai\u002Fcontactus?ref=github) 联系我们。","# EasyOCR 快速上手指南\n\nEasyOCR 是一个开箱即用的 OCR（光学字符识别）工具，支持 80+ 种语言（包括简体中文、繁体中文、英文等），能够识别拉丁文、汉字、阿拉伯文等多种书写脚本。\n\n## 环境准备\n\n*   **操作系统**：Windows, Linux, macOS\n*   **Python 版本**：建议 Python 3.8+\n*   **深度学习框架**：PyTorch\n*   **硬件要求**：\n    *   **GPU 模式（推荐）**：需要 NVIDIA 显卡及对应的 CUDA 环境，识别速度更快。\n    *   **CPU 模式**：无需显卡，可在任何机器上运行，但速度较慢。\n\n> **Windows 用户特别注意**：\n> 在安装 EasyOCR 之前，请务必先根据官方指引安装 `torch` 和 `torchvision`。\n> 请访问 [PyTorch 官网](https:\u002F\u002Fpytorch.org) 获取安装命令。\n> *   若有 GPU：请选择对应的 CUDA 版本。\n> *   若仅使用 CPU：请选择 `CUDA = None`。\n\n## 安装步骤\n\n### 1. 安装稳定版\n使用 pip 直接安装最新稳定版本：\n\n```bash\npip install easyocr\n```\n\n### 2. 安装开发版（可选）\n如需体验最新功能（可能包含未发布的修复）：\n\n```bash\npip install git+https:\u002F\u002Fgithub.com\u002FJaidedAI\u002FEasyOCR.git\n```\n\n### 3. 国内加速方案（推荐）\n在中国大陆地区，建议使用国内镜像源加速安装：\n\n```bash\npip install easyocr -i https:\u002F\u002Fpypi.tuna.tsinghua.edu.cn\u002Fsimple\n```\n\n## 基本使用\n\n### 1. Python 代码调用\n\n以下是最简单的使用示例，识别图片中的简体中文和英文：\n\n```python\nimport easyocr\n\n# 初始化 Reader，加载模型到内存（只需运行一次）\n# ['ch_sim', 'en'] 表示同时支持简体中文和英文\nreader = easyocr.Reader(['ch_sim', 'en']) \n\n# 执行识别\nresult = reader.readtext('chinese.jpg')\n\n# 打印结果\nprint(result)\n```\n\n**输出说明**：\n返回结果为列表，每项包含 `[边界框坐标，识别文本，置信度]`。\n```python\n[([[189, 75], [469, 75], [469, 165], [189, 165]], '愚园路', 0.3754989504814148),\n ([[226, 170], [414, 170], [414, 220], [226, 220]], 'Yuyuan Rd.', 0.8261902332305908)]\n```\n\n**简化输出模式**：\n如果只需要提取文字内容，不需要坐标和置信度，可设置 `detail=0`：\n\n```python\nresult = reader.readtext('chinese.jpg', detail=0)\n# 输出：['愚园路', 'Yuyuan Rd.', ...]\n```\n\n**无 GPU 环境运行**：\n如果没有显卡或显存不足，可强制使用 CPU 模式：\n\n```python\nreader = easyocr.Reader(['ch_sim', 'en'], gpu=False)\n```\n\n> **提示**：首次运行时会自动下载对应语言的模型权重至 `~\u002F.EasyOCR\u002Fmodel` 目录，请确保网络畅通。\n\n### 2. 命令行调用\n\n也可以直接在终端运行命令进行识别：\n\n```shell\neasyocr -l ch_sim en -f chinese.jpg --detail=1 --gpu=True\n```\n\n*   `-l`: 指定语言列表\n*   `-f`: 指定图片文件路径\n*   `--detail`: 是否输出详细信息 (0 或 1)\n*   `--gpu`: 是否启用 GPU","某跨境电商运营团队需要每日处理数百张包含中文、英文及阿拉伯文的多语言商品包装图，以提取关键信息录入库存系统。\n\n### 没有 EasyOCR 时\n- **多语言支持割裂**：团队需分别部署针对不同语种的识别引擎，无法在单次调用中同时处理混合了拉丁字母、汉字和阿拉伯文的复杂图片。\n- **开发门槛高且耗时**：工程师需花费数周时间预处理训练数据、调整深度学习模型参数并编写复杂的图像矫正代码，难以快速上线。\n- **环境配置繁琐**：在不同操作系统（尤其是 Windows）上配置 CUDA 依赖和视觉库极易出错，导致新成员搭建开发环境往往耗费一整天。\n- **维护成本高昂**：自研或拼凑的脚本稳定性差，遇到倾斜文字或模糊背景时识别率骤降，需人工反复复核修正。\n\n### 使用 EasyOCR 后\n- **开箱即用的多语种能力**：仅需一行代码 `reader = easyocr.Reader(['ch_sim', 'en', 'ar'])`，即可精准提取同一张图片中 80 多种语言的混合文本。\n- **极速集成与部署**：通过 `pip install easyocr` 几分钟内完成安装，直接传入图片路径或 OpenCV 对象即可获得带置信度的结构化结果，项目周期从数周缩短至数小时。\n- **跨平台兼容性强**：自动处理底层 PyTorch 依赖，无论是在 Linux 服务器还是本地 Windows 笔记本上，均能稳定运行，大幅降低环境调试痛苦。\n- **输出灵活可控**：支持一键切换“详细模式”（含坐标框与置信度）或“简洁模式”（仅文本），便于后续程序直接清洗入库，显著减少人工干预。\n\nEasyOCR 将复杂的多语言光学字符识别任务简化为几行代码，让开发者能专注于业务逻辑而非底层算法调优。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FJaidedAI_EasyOCR_392e2f04.png","JaidedAI","Jaided AI","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002FJaidedAI_f95a7b2f.png","Distribute the benefits of AI to the world",null,"https:\u002F\u002Fjaided.ai","https:\u002F\u002Fgithub.com\u002FJaidedAI",[80,84,88,92,96,100,104,108],{"name":81,"color":82,"percentage":83},"Python","#3572A5",77.1,{"name":85,"color":86,"percentage":87},"C++","#f34b7d",13.7,{"name":89,"color":90,"percentage":91},"Cuda","#3A4E3A",7.1,{"name":93,"color":94,"percentage":95},"Jupyter Notebook","#DA5B0B",1.1,{"name":97,"color":98,"percentage":99},"C","#555555",0.6,{"name":101,"color":102,"percentage":103},"Ruby","#701516",0.2,{"name":105,"color":106,"percentage":107},"Dockerfile","#384d54",0.1,{"name":109,"color":110,"percentage":111},"Shell","#89e051",0,29235,3550,"2026-04-06T02:07:52","Apache-2.0","Linux, macOS, Windows","非必需。支持 CPU 模式 (gpu=False)。若使用 GPU，需安装与系统匹配的 CUDA 版本的 PyTorch（具体版本需参考 pytorch.org），未明确指定显存大小要求，但提到低显存时可切换至 CPU 模式。","未说明",{"notes":120,"python":118,"dependencies":121},"Windows 用户必须先手动安装 torch 和 torchvision，并根据硬件选择正确的 CUDA 版本（若无 GPU 则选 CUDA=None）。模型权重首次运行时会自动下载，也可手动下载至 '~\u002F.EasyOCR\u002Fmodel' 目录。支持通过 Docker 部署。",[122,123,124,125],"torch","torchvision","numpy","opencv-python (隐含)",[16,127,14,15],"其他",[129,130,131,132,133,134,135,136,137,138,139,140,141,142,143],"ocr","deep-learning","crnn","pytorch","lstm","machine-learning","scene-text","scene-text-recognition","optical-character-recognition","cnn","data-mining","image-processing","python","easyocr","information-retrieval","2026-03-27T02:49:30.150509","2026-04-06T15:55:57.210390",[147,152,157,162,167,171],{"id":148,"question_zh":149,"answer_zh":150,"source_url":151},19417,"如何请求添加新的语言支持？","若要请求新语言支持，需要提交一个包含以下两个文件的 Pull Request (PR)：\n1. 在 `easyocr\u002Fcharacter` 文件夹中，创建名为 `你的语言代码_char.txt` 的文件，包含该语言的所有字符列表（参考该文件夹中其他文件的格式）。\n2. 在 `easyocr\u002Fdict` 文件夹中，创建名为 `你的语言代码.txt` 的文件，包含该语言的单词列表。","https:\u002F\u002Fgithub.com\u002FJaidedAI\u002FEasyOCR\u002Fissues\u002F91",{"id":153,"question_zh":154,"answer_zh":155,"source_url":156},19418,"为什么导入 easyocr 时出现 'module has no attribute Reader' 错误？","这通常是因为你将当前的 Python 脚本文件命名为了 `easyocr.py`。当代码执行 `import easyocr` 时，Python 会优先导入当前目录下的 `easyocr.py` 文件而不是安装的库模块。\n解决方案：将你的脚本文件重命名为其他名称（例如 `test_ocr.py`），并删除同目录下生成的 `easyocr.pyc` 文件或 `__pycache__` 文件夹，然后重新运行即可。","https:\u002F\u002Fgithub.com\u002FJaidedAI\u002FEasyOCR\u002Fissues\u002F122",{"id":158,"question_zh":159,"answer_zh":160,"source_url":161},19419,"初始化 Reader 时遇到模型下载失败、MD5 校验不匹配或 403 Forbidden 错误怎么办？","这通常是由于官方服务器限制了大量下载流量或网络连接问题导致的。\n解决方案：\n1. 维护者已尝试将模型托管至 CDN，如果仍然失败，可以尝试手动下载模型文件。\n2. 社区建议参考 spaCy 的做法，从 GitHub Releases 资产或通过 Google Drive 下载模型。你可以使用 `gdown` 库（`pip install gdown`）从 Google Drive 链接下载模型文件，并将其放置到 EasyOCR 的模型存储目录中。","https:\u002F\u002Fgithub.com\u002FJaidedAI\u002FEasyOCR\u002Fissues\u002F191",{"id":163,"question_zh":164,"answer_zh":165,"source_url":166},19420,"同时使用阿拉伯语和英语识别时，输出结果混乱或不正确怎么办？","目前 EasyOCR 在处理混合语言（如阿拉伯语和英语）的行分隔符方面支持有限，直接组合语言可能导致识别顺序错乱。\n解决方案：维护者建议不要依赖自动的行分隔功能，而是通过分析返回列表中的单个项目（individual items in list format）来手动处理结果。你可以分别获取每个文本块的坐标和内容，然后根据坐标逻辑自行重组或过滤多语言混合的行。","https:\u002F\u002Fgithub.com\u002FJaidedAI\u002FEasyOCR\u002Fissues\u002F25",{"id":168,"question_zh":169,"answer_zh":170,"source_url":151},19421,"EasyOCR 支持哪些主要语系的文字识别？","EasyOCR 支持多种语系，主要包括：\n- 阿拉伯脚本组：阿拉伯语、维吾尔语、波斯语、乌尔都语等。\n- 拉丁脚本组：塞尔维亚语（拉丁）、奥克语等。\n- 天城文组：印地语、马拉地语、尼泊尔语等。\n- 西里尔脚本组：俄语、保加利亚语、乌克兰语、蒙古语等。\n- 其他独立字符组：泰米尔语、希伯来语、马拉雅拉姆语、孟加拉语、旁遮普语等。\n- 东亚及改进版：日语、中文（简体\u002F繁体）、韩语的第二代模型（支持竖排文本）。\n具体支持状态可参考项目内的开发追踪列表。",{"id":172,"question_zh":173,"answer_zh":174,"source_url":156},19422,"安装后运行提示找不到 'easyocr' 模块，但 pip list 显示已安装，如何解决？","除了文件名冲突外，这可能是因为 PyTorch 版本与环境不兼容（特别是 GPU 版本驱动问题）。\n解决方案：\n1. 首先确保没有将脚本命名为 `easyocr.py`。\n2. 如果问题依旧，尝试卸载当前的 PyTorch，并安装与你的环境匹配的 CPU 版本或正确的 GPU 版本。有用户反馈在安装不带 GPU 支持的 PyTorch 后问题解决。命令示例：`pip uninstall torch torchvision` 然后访问 pytorch.org 获取适合你环境的安装命令。",[176,181,186,191,196,201,206,211,216,221,226,231,236,241,246,251,256,261,266,271],{"id":177,"version":178,"summary_zh":179,"released_at":180},117467,"v1.7.2","- 2024年9月24日 - 版本 1.7.2\n    - 修复多个兼容性问题","2024-09-24T11:24:36",{"id":182,"version":183,"summary_zh":184,"released_at":185},117468,"v1.7.1","- 2023年9月4日 - 版本 1.7.1\n    - 修复多个兼容性问题","2023-09-04T11:53:04",{"id":187,"version":188,"summary_zh":189,"released_at":190},117469,"v1.7.0","- 2023年5月25日 - 版本 1.7.0\n    - 添加对 Apple Silicon 的支持（感谢 [@rayeesoft](https:\u002F\u002Fgithub.com\u002Frayeesoft) 和 [@ArtemBernatskyy](https:\u002F\u002Fgithub.com\u002FArtemBernatskyy)，详见 [PR](https:\u002F\u002Fgithub.com\u002FJaidedAI\u002FEasyOCR\u002Fpull\u002F1004)）\n    - 修复多个兼容性问题","2023-05-25T09:11:38",{"id":192,"version":193,"summary_zh":194,"released_at":195},117470,"v1.6.2","- 2022年9月15日 - 版本 1.6.2\n    - 增加对 DBnet 检测器的 CPU 支持\n    - DBnet 只有在用户使用 DBnet 检测器初始化 EasyOCR 时才会被编译。","2022-09-15T11:34:18",{"id":197,"version":198,"summary_zh":199,"released_at":200},117471,"v1.6.1","- 2022年9月1日 - 版本 1.6.1\n    - 修复 Windows 系统下 DBNET 路径错误\n    - 新增内置模型 `cyrillic_g2`。该模型现为西里尔字母的默认模型。","2022-09-01T08:23:19",{"id":202,"version":203,"summary_zh":204,"released_at":205},117472,"v1.6.0","v1.6.0\n\n- 2022年8月24日 - 版本 1.6.0\n    - 重构代码，以支持其他文本检测器。\n    - 新增检测器 `DBNET`，详情请参阅 [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2202.10304v1)。可通过以下方式初始化使用：`reader = easyocr.Reader(['en'], detect_network = 'dbnet18')`。","2022-08-24T03:50:25",{"id":207,"version":208,"summary_zh":209,"released_at":210},117473,"v1.5.0","v1.5.0\n\n- 2022年6月2日 - 版本 1.5.0\n    - 添加了用于 CRAFT 文字检测模型的训练工具（感谢 [@gmuffiness](https:\u002F\u002Fgithub.com\u002Fgmuffiness)，详情见 [PR](https:\u002F\u002Fgithub.com\u002FJaidedAI\u002FEasyOCR\u002Fpull\u002F739)）","2022-06-02T03:32:56",{"id":212,"version":213,"summary_zh":214,"released_at":215},117474,"v1.4.2","- 2022年4月9日 - 版本 1.4.2\n    - 更新依赖库（解决 OpenCV 和 Pillow 的问题）","2022-04-09T07:42:02",{"id":217,"version":218,"summary_zh":219,"released_at":220},117475,"v1.4.1","- 2021年9月11日 - 版本 1.4.1\n    - 添加训练器文件夹\n    - 添加 `readtextlang` 方法（感谢 [@arkya-art](https:\u002F\u002Fgithub.com\u002Farkya-art)，参见 [PR](https:\u002F\u002Fgithub.com\u002FJaidedAI\u002FEasyOCR\u002Fpull\u002F525)）\n    - 扩展 `rotation_info` 参数，使其支持所有可能的角度（感谢 [abde0103](https:\u002F\u002Fgithub.com\u002Fabde0103)，参见 [PR](https:\u002F\u002Fgithub.com\u002FJaidedAI\u002FEasyOCR\u002Fpull\u002F515)）","2021-09-11T09:36:28",{"id":222,"version":223,"summary_zh":224,"released_at":225},117476,"v1.4","- 2021年6月29日 - 版本 1.4\n    - 自定义识别模型的训练与使用说明（[链接](https:\u002F\u002Fgithub.com\u002FJaidedAI\u002FEasyOCR\u002Fblob\u002Fmaster\u002Fcustom_model.md)）\n    - 示例数据集（[链接](https:\u002F\u002Fwww.jaided.ai\u002Feasyocr\u002Fmodelhub)）\n    - GPU 的批处理图像推理功能（感谢 [@SamSamhuns](https:\u002F\u002Fgithub.com\u002FSamSamhuns)，详见 [PR](https:\u002F\u002Fgithub.com\u002FJaidedAI\u002FEasyOCR\u002Fpull\u002F458)）\n    - 竖排文本支持（感谢 [@interactivetech](https:\u002F\u002Fgithub.com\u002Finteractivetech)）。此功能适用于旋转文本，不要与垂直方向的中文或日文文本混淆。（详见 [PR](https:\u002F\u002Fgithub.com\u002FJaidedAI\u002FEasyOCR\u002Fpull\u002F450)）\n    - 输出格式改为字典形式（感谢 [@A2va](https:\u002F\u002Fgithub.com\u002FA2va)，详见 [PR](https:\u002F\u002Fgithub.com\u002FJaidedAI\u002FEasyOCR\u002Fpull\u002F441)）","2021-06-29T09:44:35",{"id":227,"version":228,"summary_zh":229,"released_at":230},117477,"v1.3.2","- 30 May 2021 - Version 1.3.2\r\n    - Faster greedy decoder (thanks [@samayala22](https:\u002F\u002Fgithub.com\u002Fsamayala22)) \r\n    - Fix bug when text box's aspect ratio is disproportional (thanks [iQuartic](https:\u002F\u002Fiquartic.com\u002F) for bug report)","2021-05-30T08:55:25",{"id":232,"version":233,"summary_zh":234,"released_at":235},117478,"v1.3.1","- 24 April 2021 - Version 1.3.1\r\n    - Add support for PIL image (thanks [@prays](https:\u002F\u002Fgithub.com\u002Fprays))\r\n    - Add Tajik language (tjk)\r\n    - Update argument setting for command line\r\n    - Add `x_ths` and `y_ths` to control merging behavior when `paragraph=True`","2021-04-24T01:05:19",{"id":237,"version":238,"summary_zh":239,"released_at":240},117479,"v1.3","- 21 March 2021 - Version 1.3\r\n    - Second-generation models: multiple times smaller size, multiple times faster inference, additional characters, comparable accuracy to the first generation models.\r\n    EasyOCR will choose the latest model by default but you can also specify which model to use by passing `recog_network` argument when creating `Reader` instance.\r\n    For example, `reader = easyocr.Reader(['en','fr'], recog_network = 'latin_g1')` will use the 1st generation Latin model.\r\n    - List of all models: [Model hub](https:\u002F\u002Fwww.jaided.ai\u002Feasyocr\u002Fmodelhub)","2021-03-21T09:44:54",{"id":242,"version":243,"summary_zh":244,"released_at":245},117480,"v1.2.5","- 22 February 2021 - Version 1.2.5\r\n    - Add dynamic quantization for faster CPU inference (it is enabled by default for CPU mode)\r\n    - More sensible confident score","2021-02-23T00:29:20",{"id":247,"version":248,"summary_zh":249,"released_at":250},117481,"v1.2.4","- 7 February 2021 - Version 1.2.4\r\n    - Faster CPU inference speed by using dynamic input shape (recognition rate increases by around 100% for images with a lot of text)","2021-02-07T11:38:16",{"id":252,"version":253,"summary_zh":254,"released_at":255},117482,"1.2.3","- 1 February 2021 - Version 1.2.3\r\n    - Add `setLanguageList` method to `Reader` class. This is a convenient api for changing languages (within the same model) after creating class instance.\r\n    - Small change on text box merging. (thanks [z-pc](https:\u002F\u002Fgithub.com\u002Fz-pc), see [PR](https:\u002F\u002Fgithub.com\u002FJaidedAI\u002FEasyOCR\u002Fpull\u002F338))\r\n    - [Basic Demo on website](https:\u002F\u002Fwww.jaided.ai\u002Feasyocr)","2021-02-01T02:26:39",{"id":257,"version":258,"summary_zh":259,"released_at":260},117483,"1.2.2","- 5 January 2021 - Version 1.2.2\r\n    - Add `optimal_num_chars` to `detect` method. If specified, bounding boxes with estimated number of characters near this value are returned first. (thanks [@adamfrees](https:\u002F\u002Fgithub.com\u002Fadamfrees))\r\n    - Add `rotation_info` to `readtext` method. Allow EasyOCR to rotate each text box and return the one with the best confident score. Eligible values are 90, 180 and 270. For example, try [90, 180 ,270] for all possible text orientations. (thanks [@mijoo308](https:\u002F\u002Fgithub.com\u002Fmijoo308))\r\n    - Update [documentation](https:\u002F\u002Fwww.jaided.ai\u002Feasyocr\u002Fdocumentation).","2021-01-05T13:48:55",{"id":262,"version":263,"summary_zh":264,"released_at":265},117484,"v1.2","New language supports for Telugu and Kannada. These are experimental lite recognition models. Their file sizes are only around 7% of other models and they are ~6x faster at inference with CPU.\r\n\r\nThis release is also a preparation for user-created models\u002Farchitectures in the future.\r\n","2020-11-17T05:17:50",{"id":267,"version":268,"summary_zh":269,"released_at":270},117485,"1.1.10","- 12 October 2020 - Version 1.1.10\r\n    - Faster `beamsearch` decoder (thanks @amitbcp)\r\n    - Better code structure (thanks @susmith98)\r\n    - New language supports for Haryanvi(bgc), Sanskrit(sa) (Devanagari Script) and Manipuri(mni) (Bengari Script)\r\n- 31 August 2020 - Version 1.1.9\r\n    - Add `detect` and `recognize` method for performing text detection and recognition separately","2020-10-14T06:14:54",{"id":272,"version":273,"summary_zh":274,"released_at":275},117486,"v1.1.8","- 23 August 2020 - Version 1.1.8\r\n    - 20 new language supports for Bengali, Assamese, Abaza, Adyghe, Kabardian, Avar,\r\n    Dargwa, Ingush, Chechen, Lak, Lezgian, Tabassaran, Bihari, Maithili, Angika,\r\n    Bhojpuri, Magahi, Nagpuri, Newari, Goan Konkani\r\n    - Support RGBA input format\r\n    - Add `min_size` argument for `readtext`: for filtering out small text box","2020-08-23T04:44:29"]