A Comparative Study of Classical and Deep Learning Approaches for Bangla Handwritten Digit Recognition With Explainability

DOI

10.2991/978-94-6239-664-7_88How to use a DOI?

Keywords

Bangla handwritten digits; deep learning; traditional machine learning; transfer learning; feature engineering; Explainable AI (XAI)

Abstract

The recognition of handwritten Bangla digits are important for automating Bengali language documents in applications like postal automation, banks cheques processing and so on. Due to the morphological variation between Bangla digits and Latin digits, the automatic recognition of Bangla digit system needs special attention. This paper contrasts traditional machine learning techniques with contemporary deep learning models for Bangla handwritten digit recognition. We compare six traditional approaches with CNN-based methods using transfer learning and particularly EfficientNet-B0 and ResNet18 pre-trained weights. The Bengali Digits Dataset, containing 15,620 samples, is used for training and testing. Results demonstrate that deep learning models significantly outperform classical methods, with transfer learning achieving over 99.9% accuracy. Classical models, such as SVM and Random Forest, achieve accuracies around 85–95%. To enhance model interpretability and trustworthiness, we apply Explainable AI (XAI) techniques, including Grad-CAM, to visualize and understand model decision-making. This study establishes benchmarks for Bangla handwritten digit recognition and highlights the potential of deep learning frameworks, particularly with transfer learning and XAI, for real-world applications in low-resource environments.

Copyright

© 2026 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Mohammad Tamim Hossen
AU  - Abdulla Al Noman
AU  - Md Nafijur Rahman Nasrat
AU  - Mirza Shakil Hasan Rabby
AU  - Md. Montasir Hasan
AU  - Md. Nahid Hasan
AU  - Jahanur Biswas
PY  - 2026
DA  - 2026/06/08
TI  - A Comparative Study of Classical and Deep Learning Approaches for Bangla Handwritten Digit Recognition With Explainability
BT  - Proceedings of the International Conference on Intelligent Data Analysis and Applications (IDAA 2025)
PB  - Atlantis Press
SP  - 1301
EP  - 1316
SN  - 1951-6851
UR  - https://doi.org/10.2991/978-94-6239-664-7_88
DO  - 10.2991/978-94-6239-664-7_88
ID  - Hossen2026
ER  -