Recognizing Handwritten Offline Tamil Character by using cGAN & CNN

Abstract

The handwritten characters were being recognized by various people from different geographical locations for long period. In this, machine transcripts of human writings are important for transferring the statistics like political history, social life, financial life, religion, philosophy, and much more. Though this popularity of handwritten character recognition is achieved for a lot of languages such as English, Chinese, and Arabic, it is not achieved for Indian languages. The demanding situations explored through researchers in spotting a lot of curves, strokes, and holes in characters, massive character set, complicated letter structure, and less dataset. Generative Adversarial Network (GANs) is an interesting innovation in Deep Learning. Due to the least availability of the dataset for handwritten Tamil characters, GAN assists to boost the dataset. The enriched method referred to as GAN, is far feasible to get the specified quantity of dataset. The images can be conditionally generated by using a model generator of Conditional Generative Adversarial Network (cGAN). It is based upon a class label that permits the generation of a specific kind of image. A Convolutional Neural Network (CNN) is a class of artificial neural networks, used to recognize Tamil handwritten characters. The implementation of such methods has an accuracy of over 99 %. The proposed work would improve the accuracy by enhancing the dataset order of the Tamil Handwritten Character Recognition using cGAN and CNN.