Handwritten Ancient Tamil Character Recognition Using Generative Adversarial Network (Gan)

Abstract

Recognition Of Handwritten English Characters Has Been Done By Peoples Of Various Parts Of The Geographical Location. Although  The  Recognition  Of  These Characters  Has  Been  Achieved  For  Many  Languages  Other  Than  English,  It  Has  Not  Yet  Been  Achieved For Indian Languages Such As Tamil, Kannada, Telugu And Etc. Ancient Period Tamil Characters Are Different From Current Tamil Characters. People Have Shaped Their Discoveries In Fields Such As Medicine, Astronomy, Art And Science. The People Who Have Knowledge In Recognizing Ancient Characters Are Few. It Made The Situation To Recognize Characters That Are Written In Various Sculptures And Temples. We Will Learn About Their Lifestyle By Recognising The Letters. This Model Helps In Recognizing Tamil Characters Used In The Period Of A.D 3^rd And 4^th Century. The Aim Of The Proposed System Is To Recognize Handwritten Ancient Period Tamil Characters Using Gan Based Convolutional Neural Networks (Cnn). Generative Adversarial Networks (Gans) Are An Exciting Recent Innovation In Deep Learning. Gan Contains Generative Model Which Create New Data Instances That Resemble The Training Data. Convolution Neural Network Is Deep Learning Models Which Takes Input Generated By Gan And Process It With Its Multiple Layers To Recognize The Characters. The Implementation Of Such Training Model Results With The Accuracy Of 96 %.