Leveraging YOLO Models for Safety Equipment Detection on Construction Sites

Abstract

Occupational safety encompasses a range of practices adopted to protect the health and safety of employees. In the construction and industrial sectors, employees may be exposed to various risks such as falls, impacts, temperature changes and the effects of chemical substances. For this reason, personal protective equipment (PPE) is an important element for protecting employees against risks. The effective use of equipment such as a hardhat, mask, and vest makes an important contribution to the prevention of occupational accidents and health problems by ensuring the safety of employees. This study conducted three separate experiments investigating the potential of deep learning methods on occupational safety. In the first experiment, the YOLOv5n and YOLOv8n models were trained on the same data set with ten classes, and their performance was compared. In the second experiment, the YOLOv8n model was trained on a 2-class dataset to examine how the number of classes affected the model's performance. As a result of the experiments, it was seen that it emphasized the potential of deep learning and object detection methods to quickly and effectively monitor and evaluate the use of personal protective equipment.