Enhanced Food Classification System Using YOLO Models For Object Detection Algorithm

Abstract

Food classification is crucial for the food industry and has a significant impact on public health and nutrition. Deep learning techniques such as YOLOv5 and YOLOv7 have recently shown significant improvement in food categorization accuracy. The results of this study’s analysis of both models is evaluated on a manually labeled dataset, evaluating their accuracy, recall, and mAP at the different intersections over union (IoU) criteria. The accuracy, recall, and mAP of both YOLOv5 and YOLOv7 were evaluated at various IoU criteria, and the results showed that YOLOv5 outperformed YOLOv7 in all three metrics. Specifically, YOLOv5 achieved an accuracy score of 0.851 and a recall score of 0.836, whereas YOLOv7 scored 0.67 and 0.675, respectively. Furthermore, YOLOv5’s mAP score at the 0.5 IoU threshold was 0.892, while YOLOv7 scored 0.669. At the 0.5-0.95 IoU threshold, YOLOv5’s mAP was 0.644 and YOLOv7’s was 0.426. This study demonstrates that YOLOv5 is a superior deep learning model for food categorization tasks, with potential implications for improving consumer nutrition and food safety in the food industry.