Leveraging YOLOv8 and FaceNet for Enhanced Surveillance and Organization in Medicinal Warehouses

Abstract

Medicinal warehouses face numerous challenges, including disorganization, theft, and human errors in inventory management, which result in economic losses and unauthorized access to controlled substances. To address these issues, this study presents an advanced system that integrates the "You Only Look Once (YOLO)" algorithm for real-time object detection and tracking, implemented on a Jetson Nano microprocessor with cameras and sensors. The system automates inventory management by organizing medicinal products, reducing human error, and monitoring access to controlled substances through ultrasonic sensors and theft detection mechanisms. Additionally, a machine learning model trained on historical inventory data predicts future stock demands, enabling proactive restocking to prevent shortages and maintain operational efficiency. YOLOv8 was selected for its superior accuracy and speed, achieving a mean Average Precision (mAP) of 0.923, alongside reductions in Box Loss, Class Loss, and Distribution Focal Loss (DFL). The results demonstrate high-accuracy detection and classification, even under challenging conditions, ensuring a reliable and scalable solution for medicinal warehouse management. This system provides enhanced security, operational efficiency, and predictive capabilities, contributing to better inventory control and public health outcomes. Future work will focus on improving system scalability, integrating real-time data streams, and refining predictive accuracy for broader adoption.