Farsi Optical Character Recognition Using a Transformer-based Model

Optical Character Recognition (OCR) techniques have made significant advances in recent years using new approaches such as Transformers for Latin languages. However, research on low-resource languages like Farsi is still relatively scarce. One of the reasons for this is the complex nature of Farsi script, which poses unique challenges for OCR. Farsi OCR is essential for various applications, such as document management, digital archiving, and automated data entry. In the context of this investigation, we presents a deep neural network that utilizes transformer architecture for the recognition of Farsi words, yielding promising results. Additionally, we assess the effectiveness of our method through a comparison with state-of-the-art techniques on two datasets namely, Shotor and Sadri, achieving accuracy rates of 99.75% and 99.23%, respectively. Our results outperform previous methods and highlight the potential of transformer-based approaches for Farsi OCR.