Exploring Deep Learning Paradigms for Handwritten Digit Recognition: Insights and Advances
Amit Kumar Gupta, Priya Mathur, Nysa Maheshwari, Ayushi Mathur · 2025
In the application area of postal automation to signature verification, the Handwritten digit recognition (HRD) is major challenging field in the era of pattern recognition and machine learning. The convolutional neural networks (CNNs) becoming most emerging technology in the field of deep learning. The CNN made significant advancement in landscape of digit recognition, surpassing traditional methods in accuracy and performance. Motivated by the significance of accurate digit classification, this research presents a comprehensive review and synthesis of seminal works in the field. Through a meticulous examination of pioneering studies, we identify trends, challenges, and emerging paradigms characterizing contemporary research in handwritten digit recognition. Five deep learning methods, including a Baseline MLP, Simple CNN Model, Larger CNN Model, Basic ResNet Architecture, and Inception V3, are evaluated for their performance in digit recognition tasks. Results demonstrate that the Inception V3 model achieves the highest accuracy of 0.99, with a precision of 0.9929 and a recall of 0.9995. The review encompasses foundational works laying the groundwork for digit recognition algorithms, as well as cutting-edge research leveraging transfer learning, hybrid models, and advanced deep learning architectures. By synthesizing insights from these studies, our aim is to provide researchers and practitioners with a comprehensive understanding of state-of-art techniques, enabling informed decision-making and inspiring future advancements in HRD.