Hyderabad, 16/1 : In a significant leap for food technology tailored to the Global South, researchers at the International Institute of Information Technology Hyderabad (IIIT-H) have today unveiled a pioneering Artificial Intelligence framework designed to decode the complexity of Indian cuisine. Unlike standard Western food-tracking apps that struggle with the intricate, overlapping components of a traditional thali, this new computer vision system promises to revolutionize nutrition tracking and culinary archival.
The developments, highlighted in reports published earlier today, stem from the Center for Visual Information Technology (CVIT) at IIIT-H. The research team, led by Professor C.V. Jawahar along with authors Yash Arora and Aditya Arun, addressed a long-standing gap in AI food recognition. While existing tools easily identify discrete items like burgers or sandwiches, they fail when nuanced with the mixed textures of dal, curries, and breads found on an Indian plate.
The team has introduced a zero-shot learning system, a technical breakthrough that allows the AI to recognize food items it has never explicitly seen before without needing to be retrained. Instead of relying on a rigid classification list, the system identifies food regions on a plate and uses retrieval-based prototype matching. This flexibility is crucial for real-world deployment in diverse settings like hospital canteens or cafeterias where menus change daily.
A primary driver for this project was a healthcare mandate to monitor nutritional intake for pregnant women, requiring accurate assessment of macro-nutrients from images. The current prototype operates via an overhead camera kiosk, capable of estimating the contents and nutritional value of a meal with high accuracy. Researchers confirmed today that they are now working to transition this technology into a mobile application, having collected a vast dataset of food images from multiple angles to support phone-based scanning.
Beyond nutrition, the project holds deep cultural implications. The AI has been trained to distinguish between 12 distinct varieties of Biryani—from Hyderabadi to Kolkata and Ambur—by analyzing visual cues in cooking processes and final presentation. This effort is part of a broader vision to create an Indian Food Map, which will visualize regional variations in ingredients and cooking styles, effectively preserving the nation’s intangible culinary heritage through digital intelligence.
#IIITH #AIforGood #IndianFoodTech #ComputerVision #HyderabadInnovation #CulinaryAI
