Text Recognition in Indic Scripts using Deep Learning - A Comprehensive Analysis

The research on offline handwritten Indic scripts is still in its cradle stage, so significant research is still required in this area, whereas extensive work is already reported on Non-Indic scripts. This paper aims to discuss the advancements noted over the last few years in the area of offline handwritten Indic identification by studying different existing research works. The most key part of this survey is to represent the reported research on offline HW Indic scripts like Devanagari, Bengali, Gurumukhi etc. In this paper we presented a detailed survey of varied feature extraction as well as classification methodologies for an identification of offline HW Indic scripts. It also reports available data corpuses and the identification accuracy of the different models depending on the size of data corpus. After reporting the performed works, an analysis is done on the basis of the findings performed using different Deep Learning models.