Lexicon and attention based handwritten text recognition system

The handwritten text recognition problem is widely studied by the researchers of computer vision community due to its scope improvement and applicability daily lives. It a sub-domain pattern recognition. Due advancement computational power computers since last few decades neural networks based systems heavily contributed towards providing state-of-the-art recognizers. In same direction, we have taken two state-of-the art merged attention mechanism with it. technique has been used in domain machine translations automatic speech now being implemented domain. this study, are able achieve 4.15% character error rate 9.72% word on IAM dataset, 7.07% 16.14% GW dataset after merging beam search decoder existing Flor et al. architecture. To analyse further, also system similar Shi network greedy observed 23.27% from base model.