HANDWRITTEN TEXT RECOGNITION

Abstract

O ine handwritten text recognition from images is a signi cant is- sue for the organisa- tions endeavoring to digitize huge volumes of handmarked scanned. Handwriting recogni- tion is the capability of the computers to get and translate comprehensible handwritten in- put from sources for example paper reports, photos, contact screens and di erent gadgets into a digital format so that it can be used by computers for various purposes. In past there are various other tech- niques are used such as manual feature extraction, Hidden markov model etc. But these such techniques either requires substantially more development time or are not as much accurate. In this thesis, we created a neural network that is trained on word-pictures which is taken from the IAM dataset to translate word images into digital format.