HANDWRITTEN TAMIL CHARACTER RECOGNITION AND DIGITALIZATION USING DEEP LEARNING

Abstract

Tamil is an Ancient language predominantly spoken by Southern parts of India and also in some parts of Sri Lanka and it isthe official language of the state Tamil Nadu and the Union territory Pondicherry. Handwritten character recognition plays a vital role in conversion of written scripts into electronic forms for preserving the documents. Machine transcription of human writings could be a vital issue for transferring information relating to political history, social life, economic life, faith and philosophy. We are dealing only human handwritten characters not inscription. The ease of digitalization are quick searching, minimized storage space, fast retrieval, easy to modify, deducing new information from existing data etc. The challenges explored by researchers in recognizing Tamil characters are curves in character, large character set, complex letter structure, significant variation in writing styles due to increased number of strokes and holes, extreme font variability, difficulties faced in viewing angles, shadows and unique fonts. This work is an attempt to digitalize handwritten Tamil characters using Deep Learning techniques. This work has attained a Validation accuracy of 93% for the dataset obtained from HPLabs.