base on 带带弟弟 通用验证码识别OCR pypi版 # DdddOcr 带带弟弟OCR通用验证码离线本地识别SDK免费开源版 DdddOcr，其由 [本作者](https://github.com/sml2h3) 与 [kerlomz](https://github.com/kerlomz) 共同合作完成，通过大批量生成随机数据后进行深度网络训练，本身并非针对任何一家验证码厂商而制作，本库使用效果完全靠玄学，可能可以识别，可能不能识别。 DdddOcr、最简依赖的理念，尽量减少用户的配置和使用成本，希望给每一位测试者带来舒适的体验 项目地址： [点我传送](https://github.com/sml2h3/ddddocr) <!-- PROJECT SHIELDS --> [![Contributors][contributors-shield]][contributors-url] [![Forks][forks-shield]][forks-url] [![Stargazers][stars-shield]][stars-url] [![Issues][issues-shield]][issues-url] [![MIT License][license-shield]][license-url] <!-- PROJECT LOGO --> <br /> <p align="center"> <a href="https://github.com/sml2h3/ddddocr/"> <img src="https://cdn.wenanzhe.com/img/logo.png!/crop/700x500a400a500" alt="Logo"> </a> <p align="center"> 一个容易使用的通用验证码识别python库 <br /> <a href="https://github.com/sml2h3/ddddocr/"><strong>探索本项目的文档 »</strong></a> <br /> <br /> · <a href="https://github.com/sml2h3/ddddocr/issues">报告Bug</a> · <a href="https://github.com/sml2h3/ddddocr/issues">提出新特性</a> </p> </p> ## 目录 - [赞助合作商](#赞助合作商) - [上手指南](#上手指南) - [环境支持](#环境支持) - [安装步骤](#安装步骤) - [文件目录说明](#文件目录说明) - [项目底层支持](#项目底层支持) - [使用文档](#使用文档) - [基础ocr识别能力](#i-基础ocr识别能力) - [目标检测能力](#ii-目标检测能力) - [滑块检测](#ⅲ-滑块检测) - [OCR概率输出](#ⅳ-ocr概率输出) - [自定义OCR训练模型导入](#ⅴ-自定义ocr训练模型导入) - [版本控制](#版本控制) - [相关推荐文章or项目](#相关推荐文章or项目) - [作者](#作者) - [捐赠](#捐赠) - [Star历史](#Star历史) ### 赞助合作商 | | 赞助合作商 | 推荐理由 | |------------------------------------------------------------|------------|--------------------------------------------------------------------------------------------------| |  | [YesCaptcha](https://yescaptcha.com/i/NSwk7i) | 谷歌reCaptcha验证码 / hCaptcha验证码 / funCaptcha验证码商业级识别接口 [点我](https://yescaptcha.com/i/NSwk7i) 直达VIP4 | |  | [超级鹰](https://www.chaojiying.com/) | 全球领先的智能图片分类及识别商家，安全、准确、高效、稳定、开放，强大的技术及校验团队，支持大并发。7*24h作业进度管理 | |  | [Malenia](https://malenia.iinti.cn/malenia-doc/) | Malenia企业级代理IP网关平台/代理IP分销软件 | | 雨云VPS | [注册首月5折](https://www.rainyun.com/ddddocr_) | 浙江节点低价大带宽，100M每月30元 | ### 上手指南 ###### 环境支持 | 系统 | CPU | GPU | 最大支持py版本 | 备注 | |------------------|-----|------|----------|--------------------------------------------------------------------| | Windows 64位 | √ | √ | 3.12 | 部分版本windows需要安装<a href="https://www.ghxi.com/yxkhj.html">vc运行库</a> | | Windows 32位 | × | × | - | | | Linux 64 / ARM64 | √ | √ | 3.12 | | | Linux 32 | × | × | - | | | Macos X64 | √ | √ | 3.12 | M1/M2/M3...芯片参考<a href="https://github.com/sml2h3/ddddocr/issues/67">#67</a> | ###### **安装步骤** **i. 从pypi安装** ```sh pip install ddddocr ``` **ii. 从源码安装** ```sh git clone https://github.com/sml2h3/ddddocr.git cd ddddocr python setup.py ``` **请勿直接在ddddocr项目的根目录内直接import ddddocr**，请确保你的开发项目目录名称不为ddddocr，此为基础常识。 ### 文件目录说明 eg: ``` ddddocr ├── MANIFEST.in ├── LICENSE ├── README.md ├── /ddddocr/ │ │── __init__.py 主代码库文件 │ │── common.onnx 新ocr模型 │ │── common_det.onnx 目标检测模型 │ │── common_old.onnx 老ocr模型 │ │── logo.png │ │── README.md │ │── requirements.txt ├── logo.png └── setup.py ``` ### 项目底层支持 本项目基于[dddd_trainer](https://github.com/sml2h3/dddd_trainer) 训练所得，训练底层框架位pytorch，ddddocr推理底层抵赖于[onnxruntime](https://pypi.org/project/onnxruntime/)，故本项目的最大兼容性与python版本支持主要取决于[onnxruntime](https://pypi.org/project/onnxruntime/)。 ### 使用文档 ##### i. 基础ocr识别能力 主要用于识别单行文字，即文字部分占据图片的主体部分，例如常见的英数验证码等，本项目可以对中文、英文（随机大小写or通过设置结果范围圈定大小写）、数字以及部分特殊字符。 ```python # example.py import ddddocr ocr = ddddocr.DdddOcr() image = open("example.jpg", "rb").read() result = ocr.classification(image) print(result) ``` 本库内置有两套ocr模型，默认情况下不会自动切换，需要在初始化ddddocr的时候通过参数进行切换 ```python # example.py import ddddocr ocr = ddddocr.DdddOcr(beta=True) # 切换为第二套ocr模型 image = open("example.jpg", "rb").read() result = ocr.classification(image) print(result) ``` **提示** 对于部分透明黑色png格式图片得识别支持: `classification` 方法 使用 `png_fix` 参数，默认为False ```python ocr.classification(image, png_fix=True) ``` **注意** 之前发现很多人喜欢在每次ocr识别的时候都重新初始化ddddocr，即每次都执行```ocr = ddddocr.DdddOcr()```，这是错误的，通常来说只需要初始化一次即可，因为每次初始化和初始化后的第一次识别速度都非常慢 **参考例图** 包括且不限于以下图片 <img src="https://cdn.wenanzhe.com/img/20210715211733855.png" alt="captcha" width="150"> <img src="https://cdn.wenanzhe.com/img/78b7f57d-371d-4b65-afb2-d19608ae1892.png" alt="captcha" width="150"> <img src="https://cdn.wenanzhe.com/img/%E5%BE%AE%E4%BF%A1%E5%9B%BE%E7%89%87_20211226142305.png" alt="captcha" width="150"> <img src="https://cdn.wenanzhe.com/img/%E5%BE%AE%E4%BF%A1%E5%9B%BE%E7%89%87_20211226142325.png" alt="captcha" width="150"> <img src="https://cdn.wenanzhe.com/img/2AMLyA_fd83e1f1800e829033417ae6dd0e0ae0.png" alt="captcha" width="150"> <img src="https://cdn.wenanzhe.com/img/aabd_181ae81dd5526b8b89f987d1179266ce.jpg" alt="captcha" width="150"> <br /> <img src="https://cdn.wenanzhe.com/img/2bghz_b504e9f9de1ed7070102d21c6481e0cf.png" alt="captcha" width="150"> <img src="https://cdn.wenanzhe.com/img/0000_z4ecc2p65rxc610x.jpg" alt="captcha" width="150"> <img src="https://cdn.wenanzhe.com/img/2acd_0586b6b36858a4e8a9939db8a7ec07b7.jpg" alt="captcha" width="150"> <img src="https://cdn.wenanzhe.com/img/2a8r_79074e311d573d31e1630978fe04b990.jpg" alt="captcha" width="150"> <img src="https://cdn.wenanzhe.com/img/aftf_C2vHZlk8540y3qAmCM.bmp" alt="captcha" width="150"> <img src="https://cdn.wenanzhe.com/img/%E5%BE%AE%E4%BF%A1%E6%88%AA%E5%9B%BE_20211226144057.png" alt="captcha" width="150"> ##### ii. 目标检测能力 主要用于快速检测出图像中可能的目标主体位置，由于被检测出的目标不一定为文字，所以本功能仅提供目标的bbox位置 **（在⽬标检测⾥，我们通常使⽤bbox（bounding box，缩写是 bbox）来描述⽬标位置。bbox是⼀个矩形框，可以由矩形左上⻆的 x 和 y 轴坐标与右下⻆的 x 和 y 轴坐标确定）** 如果使用过程中无需调用ocr功能，可以在初始化时通过传参`ocr=False`关闭ocr功能，开启目标检测需要传入参数`det=True` ```python import ddddocr import cv2 det = ddddocr.DdddOcr(det=True) with open("test.jpg", 'rb') as f: image = f.read() bboxes = det.detection(image) print(bboxes) im = cv2.imread("test.jpg") for bbox in bboxes: x1, y1, x2, y2 = bbox im = cv2.rectangle(im, (x1, y1), (x2, y2), color=(0, 0, 255), thickness=2) cv2.imwrite("result.jpg", im) ``` **参考例图** 包括且不限于以下图片 <img src="https://cdn.wenanzhe.com/img/page1_1.jpg" alt="captcha" width="200"> <img src="https://cdn.wenanzhe.com/img/page1_2.jpg" alt="captcha" width="200"> <img src="https://cdn.wenanzhe.com/img/page1_3.jpg" alt="captcha" width="200"> <img src="https://cdn.wenanzhe.com/img/page1_4.jpg" alt="captcha" width="200"> <br /> <img src="https://cdn.wenanzhe.com/img/result.jpg" alt="captcha" width="200"> <img src="https://cdn.wenanzhe.com/img/result2.jpg" alt="captcha" width="200"> <img src="https://cdn.wenanzhe.com/img/result4.jpg" alt="captcha" width="200"> ##### Ⅲ. 滑块检测 本项目的滑块检测功能并非AI识别实现，均为opencv内置算法实现。可能对于截图党用户没那么友好~，如果使用过程中无需调用ocr功能或目标检测功能，可以在初始化时通过传参`ocr=False`关闭ocr功能或`det=False`来关闭目标检测功能 本功能内置两套算法实现，适用于两种不同情况，具体请参考以下说明 **a.算法1** 算法1原理是通过滑块图像的边缘在背景图中计算找到相对应的坑位，可以分别获取到滑块图和背景图，滑块图为透明背景图 滑块图 <img src="https://cdn.wenanzhe.com/img/b.png" alt="captcha" width="50"> 背景图 <img src="https://cdn.wenanzhe.com/img/a.png" alt="captcha" width="350"> ```python det = ddddocr.DdddOcr(det=False, ocr=False) with open('target.png', 'rb') as f: target_bytes = f.read() with open('background.png', 'rb') as f: background_bytes = f.read() res = det.slide_match(target_bytes, background_bytes) print(res) ``` 由于滑块图可能存在透明边框的问题，导致计算结果不一定准确，需要自行估算滑块图透明边框的宽度用于修正得出的bbox *提示：如果滑块无过多背景部分，则可以添加simple_target参数， 通常为jpg或者bmp格式的图片* ```python slide = ddddocr.DdddOcr(det=False, ocr=False) with open('target.jpg', 'rb') as f: target_bytes = f.read() with open('background.jpg', 'rb') as f: background_bytes = f.read() res = slide.slide_match(target_bytes, background_bytes, simple_target=True) print(res) ``` **a.算法2** 算法2是通过比较两张图的不同之处进行判断滑块目标坑位的位置 参考图a，带有目标坑位阴影的全图 <img src="https://cdn.wenanzhe.com/img/bg.jpg" alt="captcha" width="350"> 参考图b，全图 <img src="https://cdn.wenanzhe.com/img/fullpage.jpg" alt="captcha" width="350"> ```python slide = ddddocr.DdddOcr(det=False, ocr=False) with open('bg.jpg', 'rb') as f: target_bytes = f.read() with open('fullpage.jpg', 'rb') as f: background_bytes = f.read() img = cv2.imread("bg.jpg") res = slide.slide_comparison(target_bytes, background_bytes) print(res) ``` ##### Ⅳ. OCR概率输出 为了提供更灵活的ocr结果控制与范围限定，项目支持对ocr结果进行范围限定。 可以通过在调用`classification`方法的时候传参`probability=True`，此时`classification`方法将返回全字符表的概率 当然也可以通过`set_ranges`方法设置输出字符范围来限定返回的结果。 Ⅰ. `set_ranges` 方法限定返回字符返回 本方法接受1个参数，如果输入为int类型为内置的字符集限制，string类型则为自定义的字符集 如果为int类型，请参考下表 | 参数值 | 意义 | |-----|-----------------------------------| | 0 | 纯整数0-9 | | 1 | 纯小写英文a-z | | 2 | 纯大写英文A-Z | | 3 | 小写英文a-z + 大写英文A-Z | | 4 | 小写英文a-z + 整数0-9 | | 5 | 大写英文A-Z + 整数0-9 | | 6 | 小写英文a-z + 大写英文A-Z + 整数0-9 | | 7 | 默认字符库 - 小写英文a-z - 大写英文A-Z - 整数0-9 | 如果为string类型请传入一段不包含空格的文本，其中的每个字符均为一个待选词 如：`"0123456789+-x/=""` ```python import ddddocr ocr = ddddocr.DdddOcr() image = open("test.jpg", "rb").read() ocr.set_ranges("0123456789+-x/=") result = ocr.classification(image, probability=True) s = "" for i in result['probability']: s += result['charsets'][i.index(max(i))] print(s) ``` ##### Ⅴ. 自定义OCR训练模型导入 本项目支持导入来自于 [dddd_trainer](https://github.com/sml2h3/dddd_trainer) 进行自定义训练后的模型，参考导入代码为 ```python import ddddocr ocr = ddddocr.DdddOcr(det=False, ocr=False, import_onnx_path="myproject_0.984375_139_13000_2022-02-26-15-34-13.onnx", charsets_path="charsets.json") with open('test.jpg', 'rb') as f: image_bytes = f.read() res = ocr.classification(image_bytes) print(res) ``` ### 版本控制 该项目使用Git进行版本管理。您可以在repository参看当前可用版本。 ### 相关推荐文章or项目 [带带弟弟OCR，纯VBA本地获取网络验证码整体解决方案](https://club.excelhome.net/thread-1666823-1-1.html) [ddddocr rust 版本](https://github.com/86maid/ddddocr) [captcha-killer的修改版](https://github.com/f0ng/captcha-killer-modified) [通过ddddocr训练字母数字验证码模型并识别部署调用](https://www.bilibili.com/video/BV1ez421C7dB) ... 欢迎更多优秀案例或教程等进行投稿，可直接新建issue标题以【投稿】开头，附上公开教程站点链接，我会选择根据文章内容选择相对不重复或者有重点内容等进行readme展示，感谢各位朋友~ ### 作者 [email protected] <img src="https://cdn.wenanzhe.com/img/mmqrcode1640418911274.png!/scale/50" alt="wechat" width="150"> *好友数过多不一定通过，有问题可以在issue进行交流* ### 版权说明 该项目签署了MIT 授权许可，详情请参阅 [LICENSE](https://github.com/sml2h3/ddddocr/blob/master/LICENSE) ### 捐赠 （如果项目有帮助到您，可以选择捐赠一些费用用于ddddocr的后续版本维护，本项目长期维护） <img src="https://cdn.wenanzhe.com/img/zhifubao.jpg" alt="captcha" width="150"> <img src="https://cdn.wenanzhe.com/img/weixin.jpg" alt="captcha" width="150"> <!-- links --> [your-project-path]:sml2h3/ddddocr [contributors-shield]: https://img.shields.io/github/contributors/sml2h3/ddddocr?style=flat-square [contributors-url]: https://github.com/shaojintian/Best_README_template/graphs/contributors [forks-shield]: https://img.shields.io/github/forks/sml2h3/ddddocr?style=flat-square [forks-url]: https://github.com/shaojintian/Best_README_template/network/members [stars-shield]: https://img.shields.io/github/stars/sml2h3/ddddocr?style=flat-square [stars-url]: https://github.com/shaojintian/Best_README_template/stargazers [issues-shield]: https://img.shields.io/github/issues/sml2h3/ddddocr?style=flat-square [issues-url]: https://img.shields.io/github/issues/sml2h3/ddddocr.svg [license-shield]: https://img.shields.io/github/license/sml2h3/ddddocr?style=flat-square [license-url]: https://github.com/sml2h3/ddddocr/blob/master/LICENSE ### Star 历史 [](https://star-history.com/#sml2h3/ddddocr&Date) ", Assign "at most 3 tags" to the expected json: {"id":"3443","tags":[]} "only from the tags list I provide: [{"id":77,"name":"3d"},{"id":89,"name":"agent"},{"id":17,"name":"ai"},{"id":54,"name":"algorithm"},{"id":24,"name":"api"},{"id":44,"name":"authentication"},{"id":3,"name":"aws"},{"id":27,"name":"backend"},{"id":60,"name":"benchmark"},{"id":72,"name":"best-practices"},{"id":39,"name":"bitcoin"},{"id":37,"name":"blockchain"},{"id":1,"name":"blog"},{"id":45,"name":"bundler"},{"id":58,"name":"cache"},{"id":21,"name":"chat"},{"id":49,"name":"cicd"},{"id":4,"name":"cli"},{"id":64,"name":"cloud-native"},{"id":48,"name":"cms"},{"id":61,"name":"compiler"},{"id":68,"name":"containerization"},{"id":92,"name":"crm"},{"id":34,"name":"data"},{"id":47,"name":"database"},{"id":8,"name":"declarative-gui "},{"id":9,"name":"deploy-tool"},{"id":53,"name":"desktop-app"},{"id":6,"name":"dev-exp-lib"},{"id":59,"name":"dev-tool"},{"id":13,"name":"ecommerce"},{"id":26,"name":"editor"},{"id":66,"name":"emulator"},{"id":62,"name":"filesystem"},{"id":80,"name":"finance"},{"id":15,"name":"firmware"},{"id":73,"name":"for-fun"},{"id":2,"name":"framework"},{"id":11,"name":"frontend"},{"id":22,"name":"game"},{"id":81,"name":"game-engine "},{"id":23,"name":"graphql"},{"id":84,"name":"gui"},{"id":91,"name":"http"},{"id":5,"name":"http-client"},{"id":51,"name":"iac"},{"id":30,"name":"ide"},{"id":78,"name":"iot"},{"id":40,"name":"json"},{"id":83,"name":"julian"},{"id":38,"name":"k8s"},{"id":31,"name":"language"},{"id":10,"name":"learning-resource"},{"id":33,"name":"lib"},{"id":41,"name":"linter"},{"id":28,"name":"lms"},{"id":16,"name":"logging"},{"id":76,"name":"low-code"},{"id":90,"name":"message-queue"},{"id":42,"name":"mobile-app"},{"id":18,"name":"monitoring"},{"id":36,"name":"networking"},{"id":7,"name":"node-version"},{"id":55,"name":"nosql"},{"id":57,"name":"observability"},{"id":46,"name":"orm"},{"id":52,"name":"os"},{"id":14,"name":"parser"},{"id":74,"name":"react"},{"id":82,"name":"real-time"},{"id":56,"name":"robot"},{"id":65,"name":"runtime"},{"id":32,"name":"sdk"},{"id":71,"name":"search"},{"id":63,"name":"secrets"},{"id":25,"name":"security"},{"id":85,"name":"server"},{"id":86,"name":"serverless"},{"id":70,"name":"storage"},{"id":75,"name":"system-design"},{"id":79,"name":"terminal"},{"id":29,"name":"testing"},{"id":12,"name":"ui"},{"id":50,"name":"ux"},{"id":88,"name":"video"},{"id":20,"name":"web-app"},{"id":35,"name":"web-server"},{"id":43,"name":"webassembly"},{"id":69,"name":"workflow"},{"id":87,"name":"yaml"}]" returns me the "expected json"