Handwritten Digit Classification Using Deep Learning Convolutional Neural Network

Abstract

Due to the wide range of handwriting styles among individuals and the low image quality of the handwritten text, accurate handwriting detection has been a difficult challenge in computer vision. This is because static feature analysis of the text images is frequently insufficient to account for these factors. The accuracy of recognizing different handwriting patterns has recently progressively increased because of the introduction of machine learning, particularly convolutional neural networks (CNNs). In this study, a deep CNN model is created further to increase the handwritten digit dataset's recognition rate using various filter sizes. The proposed model's multi-layer deep structure includes a fully linked layer (also known as a dense layer) for classification and one convolution and activation layer for feature extraction. The proposed methodology has an average classification accuracy of up to 99.5% on the MNIST dataset.