CONVOLUTIONAL NEURAL NETWORK-BASED MODEL FOR HANDWRITTEN DIGIT RECOGNITION

Abstract

The topic of transcribed digit acknowledgment has been an open one in the field of pattern recognition for a long time. Several studies have shown that Neural Networks perform exceptionally well in data organization. The main goal of this study is to provide effective and reliable approaches for recognizing transcribed numerical data by examining various existing arrangement model. The Convolutional Neural Network (CNN) from Machine Learning can be used to recognize handwritten digits. The core structure of this research development is based on the MNIST (Modified National Institute of Standards and Technologies) database and CNN compilation. So, in order to run the model, libraries like NumPy, Pandas, TensorFlow, and Keras were employed. These are the fundamental pillars that the design is built upon. MNIST has roughly 60,000 photos of handwritten numbers ranging from 0 to 9. As a result, it is a classification model of class 10. The dataset is split into two parts: training and testing with Image representation as a 28*28 matrix with grayscale pixels in each cell. The result display of Convolutional Neural Networks (CNN) is the subject of this paper, which shows that the CNN classifier outperformed the Neural Network in terms of computing efficiency without sacrificing execution time. Notably, the combination of pre-processing and CNN reached the highest recognition rate of 99.16% in the experiment.