Enhancing Handwritten Devanagari Character Recognition via GAN-Generated Synthetic Data

Devanagari, one of the oldest scripts in India, remains prevalent in both Indian and Nepalese texts. However, there is a significant need to digitize these documents. While several efforts have been made for the recognition of the Roman script, fewer studies focus on the Devanagari script due to its inherent complexity. Furthermore, existing publicly available databases lack the diversity and variations to effectively train models for handling unseen data. Thus, a research gap exists in this area. To address the gap, this work proposes a model that utilizes Generative Adversarial Networks (GANs) to generate synthetic images. These images are then used to train Convolutional Neural Networks (CNNs). The model's performance is evaluated using metrics such as accuracy, precision, recall and f1-score.