BornoViT: A Novel Efficient Vision Transformer for Bengali Handwritten Basic Characters Classification

Handwritten Character Classification in Bengali script is a significant challenge due to the complexity and variability of the characters. The models which are used often, to classify characters are computationally expensive and data-hungry which are not suitable for the resource limited languages such as Bengali. In this experiment, we proposed a novel, efficient and lightweight Vision Transformer model that can classify Bengali handwritten basic characters & digits effectively, addressing some of the shortcomings of traditional methods. The proposed solution utilizes deep convolutional neural network (DCNN) in a more simplified mannar than in traditional DCNN with the aim of lowering the computational burden. With only 0.65 million parameters, 0.62 MB of model size and 0.16 GFLOPs, our model BornoViT is much lighter than the current state-of-the-art models, making our model more favorable for resource limited environment, which is quite necessary for Bengali handwritten character classification. Our model BornoViT was evaluated on BanglaLekha Isolated dataset achieving an accuracy of 95.77%, which significantly outperforms current state-of-the-art models in terms of efficiency. Furthermore, the model was evaluated using our own captured dataset, Bornomala, consisting of around 222 samples from different age groups, achieving accuracy of 91.51%.