Development and Accuracy Evaluation of a YOLOv4-Based Food Detection Model for Smart IoT Refrigerators

Efficient management of food stored in conventional refrigerators poses notable challenges, primarily due to the lack of advanced features required for inventory tracking. The absence of timely alerts further complicates users' efforts to monitor their food supplies, resulting in understocking, overbuying, spoilage, and wastage. To tackle these challenges, this work proposes a computer vision-based approach to track food items, implementing an intelligent inventory management system for IoT refrigerators. The goal is to reduce food wastage and enhance food- stocking efficiency. A YOLOv4 object detection model was trained on a custom dataset featuring common food items in Malaysian households. The model achieved a 0.8041 average loss, 100% mAP, and 86% average IoU during training. The trained model was subsequently deployed on a low-power single-board computer, implementing an autonomous and real-time inventory tracking system for IoT refrigerators. The system exhibited 93% accuracy, and macro- average scores of 0.94 for precision, 0.93 for true positive rate (TPR), 0.01 for false positive rate (FPR), 0.93 for F1 score, and 0.99 for true negative rate (TNR). Crucially, the system recognized low-stock events and sent alerts to users through the Telegram instant messaging platform, facilitating just-in-time restocking. This intelligent inventory management system offers a practical solution to address the limitations of conventional refrigeration systems and represents a transformative step towards sustainable food consumption.