A SAR-Based Approach to Recognize Text in Syloti Nagri

Optical Character Recognition (OCR) is a crucial technology used to convert both printed and handwritten text into machine-readable digital formats. In this paper, we present pioneering research for recognizing the Syloti Nagri text, which is a script used for writing the Sylheti language. While the automatic detection and recognition of the Nagri script are of significant importance, there is a noticeable scarcity of prior works in this area, with only a limited dataset available online. Indic scripts often exhibit complex character shapes and ligatures, making precise segmentation and recognition a challenging task. Our proposed approach to Nagri text recognition utilizes the Show, Attend, and Read (SAR) model, which addresses the challenge of intricate Nagri character shapes. By leveraging the Syloti Nagri OCR dataset available on Kaggle, we have achieved commendable results in the detection of Nagri characters. This research not only advances Nagri OCR but also holds great potential for digitizing historical scripts and preserving cultural heritage.