Overview

• MoScNet uses a Vision-Language Model to convert medieval Modi script into Devanagari and outperforms existing OCR solutions.
• The MoDeTrans dataset features over 2,000 expert-verified images spanning Shivakalin, Peshwekalin and Anglakalin eras.
• Both MoScNet and MoDeTrans are available on Hugging Face, enabling deployment in low-resource settings.
• Integration with platforms like BharatGPT, Bhashini and Digital India aligns the project with national digitization and UN SDG 11.4 goals.
• The open source framework offers a template for adapting the technology to other endangered or ancient scripts globally.