Personal protective equipment detection using artificial intelligence

Abstract

Every year, countless factory accidents occur in various countries. One of the important causes of workplace accidents is the inadvertent wearing of personal protective equipment (PPE) or the incomplete wearing of PPE. Therefore, addressing the necessity and importance of designing a smart system that can automatically monitor the integrity of PPE worn by workers in industrial environments. In this dissertation, Deep Learning (DL) framework-based YOLOv5 detection method is implemented to realize PPE detection, including safety helmets, goggles, masks, reflective clothes, and gloves. In the fusion of prelabeled datasets and self-labeled datasets, the detection objects are classified into 10 categories. Furthermore, evaluation metrics such as mean average precision (mAP), recall rate, and confusion metrics are used to realize a multi-faceted assessment of detection performance. With the number of 2923 images, the 10 classes mAP of this system reaches 82.6%, and the mAP of ”no goggles” and ”mask” achieves the highest which is 99.5%. In addition, this system can also be used in other occasions that have the same requirements for detection, such as hospitals, to ensure the personal safety of doctors and patients and avoid virus infection.