Offline handwritten Bangla character recognition using few-shot learning

Abstract

Character recognition is becoming more and more important as a result of its numerous applications. Optical character recognition has several uses for accessibility, storability, backups, and translation among other things, in the legal, healthcare, and financial sectors in the real world. This study focuses on a discussion of different character recognition methods for Bangla and other languages. A variety of handwritten character recognition methods have been developed for Bangla language. Yet, Bangla handwritten characters are difficult to recognize because of their variety, similarity, and compound characters. In a few papers, various deep learning techniques were applied to the recognition of Bangla characters. This paper’s goal is to suggest a novel strategy for recognizing Bangla characters. Considering the novelty of this field, the application of a few-shot learning approach, particularly because there has been no work on Bangla character recognition using this method. This approach is also well-suited for low-resource datasets. While other languages such as Chinese, Tamil, Urdu, and Malayalam have seen some work using few-shot learning, this remains an unexplored area for Bangla character recognition.