← Back to VOLUME 15, ISSUE 3, MARCH 2026
This work is licensed under a Creative Commons Attribution 4.0 International License.
CalorieScan: AI-Powered Food Recognition and Calorie Estimation System for Indian Diets
Mr. H.M. Gaikwad, Tupe Sakshi, Pawar Siddhi, Kale Vaishnavi, Rajbhoj Shravani
π 1 viewπ₯ 0 downloads
Abstract: The increasing prevalence of lifestyle diseases such as obesity, diabetes, and cardiovascular disorders in India has underscored the critical need for accessible and intelligent nutritional monitoring tools. Existing calorie tracking applications predominantly cater to Western dietary patterns, rendering them inadequate for the diverse and complex Indian food ecosystem characterized by regional variations, mixed preparations, and traditional cooking methods. This paper presents the design, development, and evaluation of CalorieScan, an AI-powered food recognition and calorie estimation system specifically engineered for Indian diets.
The proposed CalorieScan system is implemented as a Progressive Web Application (PWA) utilizing a modern technology stack comprising React.js for the frontend, Supabase with PostgreSQL for backend services, and Google Gemini 2.5 Flash Vision API for zero-shot food classification. The CalorieScan system employs a multi-modal data acquisition pipeline supporting camera-based food image capture via WebRTC, barcode scanning through the ZXing library, and voice-based food logging.
A structured prompt engineering methodology is employed to inject user-specific context - including allergen profiles, dietary goals, and regional preferences - into the vision model, enabling personalized nutritional analysis. A personalization engine within CalorieScan generates region-specific diet plans covering Maharashtrian, South Indian, North Indian, and Gujarati cuisines using affordable, locally available ingredients.
Experimental observations demonstrate that the CalorieScan system achieves practical utility with high user acceptance, offering a scalable and cost-effective alternative to manual food logging for the Indian population.
Keywords: Food Recognition, Calorie Estimation, Zero - Shot Learning, Vision AI, Indian Diet Planning, Human-in- the-Loop, Personalization, Progressive Web App, Gemini Vision.
The proposed CalorieScan system is implemented as a Progressive Web Application (PWA) utilizing a modern technology stack comprising React.js for the frontend, Supabase with PostgreSQL for backend services, and Google Gemini 2.5 Flash Vision API for zero-shot food classification. The CalorieScan system employs a multi-modal data acquisition pipeline supporting camera-based food image capture via WebRTC, barcode scanning through the ZXing library, and voice-based food logging.
A structured prompt engineering methodology is employed to inject user-specific context - including allergen profiles, dietary goals, and regional preferences - into the vision model, enabling personalized nutritional analysis. A personalization engine within CalorieScan generates region-specific diet plans covering Maharashtrian, South Indian, North Indian, and Gujarati cuisines using affordable, locally available ingredients.
Experimental observations demonstrate that the CalorieScan system achieves practical utility with high user acceptance, offering a scalable and cost-effective alternative to manual food logging for the Indian population.
Keywords: Food Recognition, Calorie Estimation, Zero - Shot Learning, Vision AI, Indian Diet Planning, Human-in- the-Loop, Personalization, Progressive Web App, Gemini Vision.
How to Cite:
[1] Mr. H.M. Gaikwad, Tupe Sakshi, Pawar Siddhi, Kale Vaishnavi, Rajbhoj Shravani, βCalorieScan: AI-Powered Food Recognition and Calorie Estimation System for Indian Diets,β International Journal of Advanced Research in Computer and Communication Engineering (IJARCCE), DOI: 10.17148/IJARCCE.2026.15342
This work is licensed under a Creative Commons Attribution 4.0 International License.