This model is trained on a very small sample of data. It obviously doesn't work for different kinds of captcha images. But the usage of R-CNN along with CTC loss is interesting. This project was inspired by Abhishek Thakur's YouTube video.
To train the model, run:
python train.py
To test the model on an image, add the image to the output folder and pass the file name as an argument:
python inference.py --filename=test2.png