MODEL FOR DETECTING SAFETY HELMET WEARING USING IMPROVED YOLO-M

Abstract

Construction site safety remains a critical concern, necessitating innovative solutions to ensure the well
being of workers. This study introduces an intelligent safety helmet detection system leveraging computer vision 
technology to monitor and enforce safety protocols in real-time. Through comprehensive analysis, we compare the 
performance of various state-of-the-art object detection architectures, including YOLOv5s, YOLOv5 - YOLO M, 
SSD, RetinaNet, FasterRCNN, YOLOv3, YOLOv4, YOLOv5 - GhostCNN, and YOLOv8. Our evaluation focuses 
on Recall,Presicion, Mean Average Presicion (mAp) aiming to provide insights into their suitability for safety 
compliance applications in the construction industry. The primary beneficiaries are construction workers, whose 
safety is paramount, alongside site managers who can optimize resource allocation and streamline monitoring 
efforts. Initial results demonstrate YOLOv5 - GhostCNN's potential to achieve over 97% mean Average Precision 
(mAp), suggesting promising avenues for further enhancing workplace safety. This research contributes to a safer 
working environment, facilitating better adherence to safety regulations and reducing the risk of construction-related 
accidents.