MODI Charater Recognition using Convolutional Neural Networks

Abstract

Convolutional Neural Network also known as ConVNet or CNN algorithm is a Deep learning algorithm, which automatically detects the important features in an image without human supervision. The preprocessing required for ConVnet  is much lower comparatively to other classification algorithm. The CNN successfully determines the spatial and temporal dependencies in an image through several filter. The CNN model has different layers that makes classification and feature extraction process easier. Thus CNN can be used for character recognition process. With help of Character recognition using CNN we can convert the Modi handwritten script into digitized English text and corresponding Devanagari character script. Modi script is used to write the Marathi language. Most of the ancient documents in Maharastra were written in Modi language. Hence the translation of the Modi script is necessary.In this paper the character recognition of the handwritten Modi script is performed using CNN algorithm. The pre-processing, training of CNN model are all written in python programming language in JupyterLab. The system developed not only translates the image character to the English but also display Devanagari corresponds of the Modi script predicted.The overall system accuracy is 96% for 46 character of the Modi script used.