India is a massively multilingual country, with diverse languages and scripts that present unique challenges for technology development. While rapid progress has been made in English language technologies through tools like ChatGPT and other large language models (LLMs), Indian languages remain significantly underserved. At IIT Jodhpur, the Vision, Language and Learning Group (VL2G), led by Prof. Anand Mishra from the Department of Computer Science and Engineering, is working to bridge this gap by developing indigenous AI technologies that integrate vision and language, ensuring inclusivity and accessibility for all.

Breaking Barriers in Indian Language Technologies

As part of the Government of India's flagship project Bhashini , the team has developed advanced models to read and translate Indian language text appearing in images. Unlike commercial tools that largely focus on English, the IIT Jodhpur group has already deployed open-source APIs capable of reading scene text in 13 major Indian languages including Hindi, Marathi, Gujarati, Punjabi, Bengali, Assamese, Odia, Manipuri, Malayalam, Kannada, Tamil, Telugu, and Urdu. These tools have the potential to power multilingual navigation, real-time translation of street signs, and improve accessibility for billions of Indian users.

Preserving India's Cultural Heritage through AI

In collaboration with the Indira Gandhi National Center for the Arts (IGNCA) and TIH- iHUB Drishti, the group is digitally preserving and restoring ancient manuscripts written in Sanskrit, Pali, Telugu, and other languages. Using deep learning-based computer vision techniques, the team is cleaning, enhancing, and restoring deteriorated manuscripts while developing OCR systems to digitize these priceless texts. This effort will safeguard India's cultural heritage while enabling new translations, comparative studies, and rediscovery of knowledge across disciplines.

Video AI for Skills, Safety, and Society

The group is also advancing fine-grained video understanding with real-world applications. In collaboration with Accenture Labs, one of Prof. Mishra's PhD students have developed a state-of-the-art model capable of localizing and tracking unseen objects in videos using advanced transformer networks. Such models can support automated skill assessments, workplace safety monitoring, surveillance, and video analytics across industries such as manufacturing, healthcare, and education.

Speaking about the broader vision of his research, Prof. Anand Mishra said, “Our aim is to create robust, indigenous, and open-source technologies that are accessible to all. By focusing on Indian language text recognition, cultural preservation, and advanced video understanding, we want to ensure that India not only keeps pace with global AI advancements but also contributes unique solutions rooted in our societal and cultural needs.”

Through these initiatives, IIT Jodhpur is demonstrating how cutting-edge AI research can simultaneously address India's pressing challenges and make significant contributions to global knowledge and innovation.

Video - Sparks of IIT Jodhpur I Dr. Anand Mishra I Bridging Language & Vision

Drive Link

(Release ID: 2161258) Visitor Counter : 434