Performance Analysis of Deep CNN, YOLO, and LeNet for Handwritten Digit Classification

The fundamental applications of handwritten digit classification are in the fields of optical reorganization of digits, bank check processing, recognizing zip codes on mail for postal, processing bank check amounts, and numeric entries in forms filled up by hand. For processing these types of tasks, different kinds of learning algorithms are used. A comparative study on performance analysis was done for convolutional neural network, LeNet-5, and YOLOv7. Publically available MNIST, DIDA, and MNIST MIX handwritten digit dataset were used in experimental work. The objective of this study is to find the best algorithm which can give an acceptable accuracy. To implement the model, this paper uses a deep neural network CNN and its architecture LeNet-5 and YOLOv7 which have become a potent tool for image categorization problems in recent years. This paper demonstrates the efficacy of deep learning approaches for effective and precise digit recognition, which can be expanded to numerous real-world applications needing accurate and dependable recognition of digits written on paper. This research has achieved the highest accuracy of 99.38% for LeNet-5.