Bangla Handwritten Digit Recognition Using Convolution Neural Network

Abstract

Bangla handwritten digit recognition is always a big challenge due to its variation of shape, size, and writing style. Due to the economic and educational values of accurate handwritten recognition, researchers are becoming more thoughtful about it. Several works have been already done on the Bangla Handwritten Digit Recognition. Therefore, in this paper, we investigated the dynamic search process to recognize a digit. The unbiased dataset, NumtaDB is used for Bangla digit recognition. This paper states the development and implementation of a lightweight Convolution Neural Network (CNN) model for classifying Bangla Handwriting Digits. We have systematically evaluated the performance of our method on this image database NumtaDB. And finally, from experiments, we have achieved a 99% accuracy using some of the proposed methods.