For this task, I'm going to implement a Connectionist temporal classification(CTC) based convolutional recurrent neural network(CRNN) ocr model that transcripts english handwritten images to strings.
Steps:
- Build a data pipeline
- Implement an image encoder (most likely a convolutional neural network encoder)
- Build a decoder (most likely a recurrent neural network based decoder)
- Implement the training function to train the model
- Implement the validation function to validate the trained model