Multilingual Handwritten Digit Recognition Using Multiplexer-Based Deep Learning Models
Amaan Mansuri, Atir Sakhrelia, Anaya Patel, Priyank Thakkar, Ankit K. Sharma, Krishn Limbachiya · 2024
Recognizing handwritten digits poses a significant challenge in machine learning and image processing due to the inherent variations in individual writing styles, sizes, curves, strokes, and interpretations. Using deep learning models, the research presented here describes an innovative method for classifying handwritten digits in various languages, including Urdu, Gujarati, Hindi, and Bengali. The approach proposed utilizes a model that employs a multiplexing concept to build a more efficient and scalable model as compared to a naive 50 -class classifier model capable of distinguishing between the digits. The use of a multiplexer-based approach streamlines the classification process and achieves $\mathbf{9 8. 8 8 \%}$ accuracy, offering a promising solution for recognizing handwritten digits in diverse linguistic contexts.