Pre-Trained Deep Learning Models for Detecting Strikeouts in Kannada Handwritten Documents

Abstract

In the digital document analysis, handwritten character recognition is a challenging area.  Various methods are proposed in the literature to identify the strike-outs in various languages. Kannada handwritten classification is a crucial machine vision problem due to its various practical applications in the development of recognition systems. Identifying of strikeout string on the given Kannada statement is an important challenge to develop robust recognition system. In the present study, pre-trained models such as DenseNet121, EficientNetB0, MobileNet, InceptionResNetV2 are used to recognize strikeout Kannada words.  Experimental analysis indicates that EficientNetB0 performed better for strikeout Kannada words and InceptionResNetV2 results in higher performance for non-strikeout Kannada words.