Breaking Captchas with Convolutional Neural Networks

• This project is a captcha breaker system that utilizes deep learning to achieve its goal.
• The model has been trained using LeNet architecture.
• An example of a captcha we aim to break is shown below.
  
  

The steps followed to build the model:

1. Download the captcha dataset.
2. Split dataset into training and testing datasets.
3. Labelling images. The annotate.py script annotates all of the extract the digits from each image in downloads images and hand-label every digit automatically.
4. Training the model. The train_model.py script will train LeNet on the labelled digits.
5. Testing and evaluating our model on example images. The test_model.py will apply LeNet to captcha images themselves.

Result

This model has achieved 100% accuracy. However, the used dataset might be simple.

Samples