Git Product home page Git Product logo

deeplearningslidecaptcha's Introduction

滑动验证码深度学习识别

本项目使用深度学习 YOLOV3 模型来识别滑动验证码缺口,基于 https://github.com/eriklindernoren/PyTorch-YOLOv3 修改。

只需要几百张缺口标注图片即可训练出精度高的识别模型,识别效果样例:

克隆项目

由于本项目使用了 Git LFS,包含了已经训练好的模型,克隆时间较长,基本命令:

git clone https://github.com/Python3WebSpider/DeepLearningSlideCaptcha.git

如果想加速克隆,暂时先跳过大文件模型下载,可以执行命令:

GIT_LFS_SKIP_SMUDGE=1 git clone https://github.com/Python3WebSpider/DeepLearningSlideCaptcha.git

数据准备

使用 LabelImg 工具标注自行标注一批数据,大约 200 张以上即可训练出不错的效果。

LabelImg:https://github.com/tzutalin/labelImg

标注要求:

  • 圈出验证码目标滑块区域的完整完整矩形,无需标注源滑块。
  • 目标矩形命名为 target 这个类别。
  • 建议使用 LabelImg 的快捷键提高标注效率。

环境准备

建议在 GPU 环境和虚拟 Python 环境下执行如下命令:

pip3 install -r requirements.txt

预训练模型下载

YOLOV3 的训练要加载预训练模型才能有不错的训练效果,预训练模型下载:

bash prepare.sh

下载完成之后会在 weights 文件夹下出现模型权重文件,供训练使用。

训练

本项目已经提供了标注好的数据集,在 data/captcha,可以直接使用。

如果要训练自己的数据,数据格式准备见:https://github.com/eriklindernoren/PyTorch-YOLOv3#train-on-custom-dataset

当前数据训练脚本:

bash train.sh

实测 P100 训练时长约 15 秒一个 epoch,大约几分钟即可训练出较好效果。

测试

训练完毕之后会在 checkpoints 文件夹生成 pth 文件,可直接使用模型来预测生成标注结果。

如之前跳过了 Git LFS 文件下载,则可以使用如下命令下载 Git LFS 文件:

git lfs pull

此时 checkpoints 文件夹会生成训练好的 pth 文件。

当前数据测试脚本:

sh detect.sh

该脚本会读取 captcha 下的 test 文件夹所有图片,并将处理后的结果输出到 test 文件夹。

运行结果样例:

Performing object detection:
        + Batch 0, Inference Time: 0:00:00.044223
        + Batch 1, Inference Time: 0:00:00.028566
        + Batch 2, Inference Time: 0:00:00.029764
        + Batch 3, Inference Time: 0:00:00.032430
        + Batch 4, Inference Time: 0:00:00.033373
        + Batch 5, Inference Time: 0:00:00.027861
        + Batch 6, Inference Time: 0:00:00.031444
        + Batch 7, Inference Time: 0:00:00.032110
        + Batch 8, Inference Time: 0:00:00.029131

Saving images:
(0) Image: 'data/captcha/test/captcha_4497.png'
        + Label: target, Conf: 0.99999
(1) Image: 'data/captcha/test/captcha_4498.png'
        + Label: target, Conf: 0.99999
(2) Image: 'data/captcha/test/captcha_4499.png'
        + Label: target, Conf: 0.99997
(3) Image: 'data/captcha/test/captcha_4500.png'
        + Label: target, Conf: 0.99999
(4) Image: 'data/captcha/test/captcha_4501.png'
        + Label: target, Conf: 0.99997
(5) Image: 'data/captcha/test/captcha_4502.png'
        + Label: target, Conf: 0.99999
(6) Image: 'data/captcha/test/captcha_4503.png'
        + Label: target, Conf: 0.99997
(7) Image: 'data/captcha/test/captcha_4504.png'
        + Label: target, Conf: 0.99998
(8) Image: 'data/captcha/test/captcha_4505.png'
        + Label: target, Conf: 0.99998

样例结果:

协议

本项目基于开源 GNU 协议,另外本项目不提供任何有关滑动轨迹相关模拟和 JavaScript 逆向分析方案。

本项目仅供学习交流使用,请勿用于非法用途,本人不承担任何法律责任。

如有侵权请联系个人删除,谢谢。

deeplearningslidecaptcha's People

Contributors

0asa avatar catalinolaru1 avatar eriklindernoren avatar everwinter23 avatar germey avatar id9502 avatar timforby avatar

Stargazers

 avatar  avatar

Watchers

 avatar

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.