nsysu-captcha-solver

Captcha solver across websites under selcrs.nsysu.edu.tw, National Sun Yat-sen University

Chrome extension

Todo

• Chrome extension
• Firefox extension
• Safari extension

Project Structure

.
├── README.md                       # this file
├── README.zh_tw.md                 # the Chinese version
└── src                             
    ├── extension                   # The extension folder
    │   ├── chrome
    │   │   ├── content.js          # content script
    │   │   ├── icons               # The extension icons
    │   │   ├── manifest.json       # manifest file
    │   │   └── tf.min.js           # tensorflow js package
    │   └── safari                  # TBA
    └── neural_network_engine
        ├── data_preprocessing      
        │   ├── full_len_data       # original images
        │   ├── split_img.py        # generate splitted training data and labels
        │   └── utils
        │       ├── get_img.py      # primitive image crawler
        │       ├── model.h5        
        │       └── predict_img.py  # predict single captcha image
        ├── dataset
        ├── labels.csv              # data instance's path & label
        ├── outputs                 # model outputs of train.py
        ├── train.py
        └── train_demo.ipynb

Usage

Chrome extension

Effective immediately after installation

Safari extension

Currently under construction.

Train backend CNN engine

1. Prepare your training images and place them under full_len_data
2. Use split_img.py to generate processed training data and labels.csv. The default output directory is dataset.
3. Run train.py to train the CNN, the output model will be under outputs.
4. The model this project currently using is hosted under the gh-pages branch.

Thanks

Special thanks to

• SCC For data collection
• Paul Peng For creative contents