Convert Arabic Handwritten Images to Text
Given a Arabic handwritten word in image form. Convert it into text form or recognise the word and get the word in text form
Used 3 layer CNN to learn the features of Arabic Text. Fed that leaning into dynamic_rnn module with LSTM cell to predict the output.
The model is trainable and able to converge on small dataset.
to train the model
Implemntation of CNN-LSTM model using tensorflow.
Configuration file for all the hyperparameters and variable
contains helpers function to create batch to feed into model, prepare arabic character set and dictionary
contains the results of training along with predicted result
Need good amount of GPU resource to train the model on full dataset to get the Results.