Classification of Ancient Modi Script Characters using Convolutional Neural Network

Abstract

This study aims to develop a character recognition system for the MODI language, which is challenging due to the complexity of character identification. The proposed system uses convolutional neural networks (CNN) and the VGG16 algorithm to accurately identify printed and handwritten MODI characters from scanned papers, regardless of the input paper quality. The training dataset consists of 48 distinct MODI script characters, including vowels and consonants, and is routinely updated with handwritten samples obtained from various sources and the IEEE DataPort dataset. The study demonstrates the feasibility of developing a highly accurate character recognition system that follows the established method of the MODI Script Character Recognizer System (MSCR).