A DEEP LEARNING-BASED SYSTEM FOR AUTOMATED DIETARY ASSESSMENT USING IMAGE RECOGNITION

Abstract

Automated dietary assessment is crucial in monitoring nutritional intake and promoting healthier eating habits. This study presents a deep learning-based system designed to enhance the accuracy and efficiency of dietary assessments using image recognition. The proposed system leverages convolutional neural networks (CNNs) for the automated identification and classification of food items in images captured by users. By integrating advanced image processing techniques with deep learning models, the system can accurately detect and classify a wide range of food types and portion sizes. To develop this system, a large dataset of food images was curated, including various cuisines, preparation methods, and portion sizes. The CNN model was trained and fine-tuned to recognize specific food items, accounting for visual variations due to different angles, lighting conditions, and presentation styles. The system also incorporates a portion estimation module, which calculates the nutritional content based on recognized food items and their estimated portions. This allows for a comprehensive assessment of daily dietary intake, providing users with detailed feedback on their nutritional habits. Validation of the system was conducted using a series of tests with real-world food images, demonstrating high accuracy in both food recognition and portion estimation. The results indicate that this deep learning-based system offers a promising solution for automated dietary assessment, with potential applications in personal health management, clinical nutrition, and public health monitoring. Future work will focus on expanding the food database, improving portion size estimation, and integrating the system into mobile health applications for real-time dietary tracking.