SmartBite: Where AI Meets Appetite – A Vision Based System for Real-Time Food Recognition and Caloric Intelligence

Abstract: With increasing cases of diet-related conditions, such as obesity and lifestyle diseases, the need for effective and user-friendly tools to track calorie intake has never been more urgent. Manual food logging methods often suffer from inaccuracies and are impractical for long-term use. This paper introduces SmartBite, an AI-powered application that estimates calorie content from food images using deep learning. The system leverages a fine-tuned ResNet50 CNN model trained on the Indian Food-80 dataset and features a lightweight web interface built with Flask. Users can input food either through images or text, with Google Gemini AI supporting natural language understanding. The model achieves 96% accuracy with an average response time of 2.21 seconds, making SmartBite a practical tool for fast and reliable dietary monitoring.