This is a repo containing tensorflow implementation of the paper EATEN: Entity-aware Attention for Single Shot Visual Text Extraction. This is a single shot text extraction model for extracting Entity of Interest. The data mentioned in the paper can be found here. In the paper three different datasets were used namely train tickets, business cards,passport. The code in this repo is written for a subset of train tickets dataset due to computational resources constrains,however it can be easily extended to all the three datasets which are mentioned in the paper with small changes to the code.

Clone the git repo and download the cnn encoder, entity decoders, tokenizers from here and then execute gui.py