RASOI – Recognition and Segmentation of Indian Thali using AI

Data Challenge arranged along with NCVPRIPG 2025

National Institute of Technology Srinagar

from 16th to 18th July 2025

(Jointly organized with IIT Jammu and NIT Jalandhar)

The Indian Thali is visually complex and culturally diverse. This challenge aims to push the frontiers of AI by identifying each food item in an Indian Thali.

Motivation

Food recognition plays a key role in diet tracking, calorie estimation, healthcare, and personalized nutrition. Though there exists a lot of work on other cuisines; Indian food, especially the composite Thali is underrepresented. The diversity and the uncertainty of composition of the Indian thali pose a unique vision challenge: dishes may overlap, vary regionally, and have similar textures or colors. We believe that this challenge will contribute significantly towards applications in nutrition and healthcare, personalized food recommendation systems and cultural preservation.

Objective

To develop and evaluate models that can accurately demarcate each food item and label it in an image of a typical Indian Thali.

The participants will work on the following three problem statements.

Object Detection

Bounding box localization of each food item

Food Classification

Assigning each box to its correct food category

Total Calorie Estimation

Estimating the calorie from images and portions

Registration

The participant should form a team of minimum two and maximum three members and register their team using the below Google form on or before the deadline.

Note that each team should register themselves to be eligible for this contest and get access to the data. Without registration, the team will not be considered.