I am trying to build an OCR which can read the Mathematical equations just like MAthPix and im2markup. im2markup by HarvardNLP seems like a good model but the thing is that it is built using Torch and I want to implement it using the Keras or Pytorch.
I am thinking about using a CNN most preferably reinforcement learning like VGG or ResNet for feature extraction from the images just like in the paper Show, Attend and Tell: Neural Image Caption Generation with Visual Attention which has an implementation in image captioning using Attention
Which method for building the network should give me the best results as Decoder? I believe the Attention Mechanism is supposed to give me the best results and second bet is with CRNN.
Also, is there a good pre-built model with PyTorch or Keras that I can use instead of building my own which produces satisfactory results? Please do suggest.