This is a simple OCR project which can recognize the character present in the image. For now, it is only capable of recognizing digits and capital letter.
For digits, I have used MNIST dataset, and for alphabets I have used kaggle A-Z Handwritten Alphabets
For the model architecture I have used Resnet architecture which you can read about here.
This is the image of the architecture I have used.
First install the required modules from requirements.txt
If you would like to test the model, you can run GUI_predict.py and try. Here is the sample output you could expect.
If you would like to train the model, you need to download and extract the 'A-Z Handwritten Alphabets' to data/ and edit the ocr_train.py as per your needs. I have set it to 50 epochs, which would take around 300sec/epoch in google colab