Tibetan Character Recognition Based on GOT-OCR2.0
Yao Ji, Jie Zhu, Mingzheng Zhu · 2025
This study introduces GOT-OCR2.0, a novel framework for Tibetan text recognition, leveraging the GOT model to achieve efficient and real-time character recognition. The proposed end-to-end approach addresses limitations of traditional methods in complex backgrounds and diverse fonts. Experimental results show superior performance and robustness in low-light and high-noise conditions. The paper also discusses the model's advantages, challenges, and future directions, offering technical support for Tibetan digital resource development.