Revolutionizing Heritage Preservation: IIT Roorkee's AI Breakthrough

In a landmark development, the Indian Institute of Technology (IIT) Roorkee has introduced the world's pioneering AI framework for translating Modi script into Devanagari, as confirmed by officials on Friday. This trailblazing initiative, MoScNet, harnesses a Vision-Language Model (VLM) architecture to facilitate the preservation and digitization of ancient manuscripts under projects like Digital India and Bhashini.

Titled 'Historic Scripts to Modern Vision', the project debuts 'MoDeTrans', an unprecedented dataset containing over 2,000 images of Modi script manuscripts covering three historical periods: Shivakalin, Peshwekalin, and Anglakalin, alongside expert-verified Devanagari transliterations. Headed by Sparsh Mittal of IIT Roorkee, the MoScNet model surpasses existing OCR technologies, offering a robust, lightweight solution, particularly effective in resource-constrained settings.

Students Harshal, Tanvi, and Onkar, affiliated with COEP Technological University and Vishwakarma Institute of Information Technology, Pune, also played vital roles. Kamal Kishore Pant, Director of IIT Roorkee, highlighted the groundbreaking work as a means to revitalizing cultural heritage, bolstering academic studies, and supporting nation-building. This effort, which capitalizes on AI for both transliteration and cultural preservation, aspires to democratize access to India's rich, historical records, aiding historians, researchers, and governmental archives.

(With inputs from agencies.)