Food Recognition and Calorie Estimation in Mixed Food Items using MobileNet

Artificial Intelligence (AI) and deep learning have significantly enhanced food recognition and calorie estimation, but the majority of current models are not good at mixed food items, where multiple ingredients are mixed together to form one dish. This research proposes an AI-based system for food recognition and calorie estimation in mixed food items through deep learning-based object detection, image segmentation, and volume estimation methods. Compared to traditional approaches using fixed portions and single-label classification, our approach employs Mobilenet-based multi-label detection and semantic segmentation to accurately identify individual food constituents in a meal. The system is trained using a large, heterogeneous dataset of mixed food images, with high food recognition and caloric estimation accuracy. We also employ explainable AI methods to enhance model transparency and trust. The proposed solution is implemented in an intuitive mobile application, enabling real-time food monitoring and personalized dietary insights. Future development will include increasing the dataset, enhancing portion estimation, and incorporating real-time nutritional guidance. This research attempts to bridge the gap between AI and nutrition science, providing a more accurate and efficient dietary analysis tool for consumers.