International Research Journal of Education and Technology

The ability to accurately identify and understand handwritten digits plays a crucial role in various applications, including document processing, postal automation, and banking systems. However, recognizing handwritten digits, particularly in complex scripts like Hindi, presents a significant challenge due to variations in writing styles, character ambiguities, and noise. This research explores the application of Convolutional Neural Networks (CNNs) for robust and efficient Hindi digit identification. Other applications using artificial intelligence techniques such as CNN, deep CNN to identify Hindi numeric text provide real-time quick analysis but might not be as efficient when it comes to variations in handwritten digits. This study assesses AI’s practicality in identifying Hindi digits from 0 to 9. Our model has demonstrated remarkable performance, achieving an accuracy of over 95.95% in identifying handwritten Hindi digits. While it does have a 5% error rate, indicating a slight decrease in performance, it remains manageable. The slight dip in accuracy may be attributed to the variations in handwriting styles or the similarities in linguistic strokes of Hindi language digits. This error rate could potentially be addressed through more comprehensive training of our model. To achieve this we have used a 2 layer architecture in our model.

Keywords: Hindi Handwriting Digit Recognition, CNNs, Deep Learning, Digit Recognition, Pattern Recognition, Machine Learning, Machine Learning Algorithms, Deep Learning Algorithms, Natural Language Processing (NLP), Natural Language Understanding (NLU), Hindi Digit Recognition, Vernacular Language Understanding, Confusion Matrix, ResNet, Model Training from scratch.