Dietary Management System Based on the DeepRD Model

With the development of deep learning technology, the diet of diabetic patients is recorded by intelligent means to help them control their blood glucose through scientific dietary management. This study proposes an image-based food recognition scheme to help diabetic patients control their blood sugar through deep learning techniques. A food database containing 170,000 images including 241 dishes was created for this study. A deep learning driven food recognition model (DeepRD) is proposed. The system was also trained and validated on the proposed system with a Top1 classification accuracy of 80.38 %. And for the first time, the food glycemic index is introduced into the food recognition task to help diabetic patients to control their blood glucose scientifically by managing their diet.