Smart food area–weight system for dietary and nutritional monitoring

This study proposes a diet and nutrition monitoring system that combines deep learning and IoT technologies to provide accurate and user-friendly food portion estimation. The proposed system contains scalable stationary modules, centralized server, and user interface designed for organizational dining areas. The stationary module utilizes a camera, a Bluetooth Low Energy weight sensor, and a tablet for automatic portion estimation of multiple foods placed on a tray from one image. The centralized server deploys deep neural network models for food recognition and portion estimation, providing nutrition calculations, and storing users’ dietary records and user interface is provided for users to review their dietary records and nutritional information on the online web platform and mobile application. This deep learning pipeline based on object detection and segmentation is used to identify food types and estimate quantities, with area–weight-based estimation applied in the system. Evaluation results demonstrate that the stationary module achieves high precision, with several food items showing a mean absolute error below 7 g and a mean absolute percentage error near or under 10%. Despite its slightly low accuracy, the proposed system provides a scalable, convenient, and automated approach to dietary tracking, suitable for large institutions. Copyright © 2025 IEEE.