Virtual Booth

Diabetes is a prevalent concern among Malaysian adults, requiring effective dietary intake monitoring. Existing dietary tracking methods often fail to accurately recognize Malaysian dishes. This study employs deep learning techniques to improve calorie estimation for Malaysian cuisine. It evaluates three models—MobileNetV2, VGG19, and ResNet-50—using 2,076 images of 20 Malaysian dishes. MobileNetV2 achieved the highest accuracy at 97.83%, while ResNet-50 had the lowest at 18.11%. Post-classification, YOLOv8 estimated serving quantities with an [email protected] of 88%. The system, integrating these models, was deployed in mobile and web applications, proving effective in real-world scenarios.