An empirical evaluation of proposed model based on CNN for handwritten digit recognition

Abstract

Computer vision is an emerging field of computer science that deals with visual inputs such as images and videos. It enables computers and systems to understand and process information from videos and images as processed by human beings. Computer vision is brought to life by convolutional neural networks (CNN) and deep learning. Deep learning is a subfield of machine learning that teaches computers to learn by example. CNN is an artificial neural network that enables deep learning models to break down images into pixels which are given tags and labels. Handwritten digit recognition is the process of identifying handwritten digits by computer systems. In this paper, a CNN model is used on the MNIST dataset to identify handwritten digits. The aim of this paper is to find the model that gives maximum accuracy based on various hyper parameters such as kernel size, learning rate, batch size, number of hidden layers, epochs.