CNN-LSTM-CTC Amharic OCR

End-to-End learning

• This method uses CNN-LSTM-CTC networks.
• To train and test this model, please use the ADOCR dataset from http://www.dfki.uni-kl.de/~belay/. (new link) https://bdu-birhanu.github.io/amharic.html However, to play with the code you may also use the sample database given, together with the source code, in this directory.
• Both the text-line images and corresponding ground truth are given in numpy format.

To run the code with Terminal use the following info:

# Load and Pre-process data
python data_loader.py

# Train
python train_model.py

# Test and results
python test_model.py

Some issues to know

1. The test environment is
   • Python 2.7
   • Keras 2.2.4
   • tensorflow 1.14