I am using keras_ocr for doing receipt scanning and using the SROIE 2019 receipt dataset. I am finding that the recognizer underfits the data -- the loss and val_loss plateau around 11.0. I would greatly appreciate any help given since I don't have anyone to discuss the problem with...
I am using
keras_ocr
for doing receipt scanning and using the SROIE 2019 receipt dataset. I am finding that the recognizer underfits the data -- the loss and val_loss plateau around 11.0. I would greatly appreciate any help given since I don't have anyone to discuss the problem with...Thanks, Hadil