Safety Helmet Detection Based on YOLOv7
Kequan Chen, Guibao Yan, Muqi Zhang, Zhang-Shu Xiao, Qichao Wang
Abstract
Frequent safety accidents have posed a significant risk to workers' lives recently. In particular, the risk of head injury is significantly increased by workers not wearing helmets. However, manual supervision is inefficient and costly. Even though traditional object detection methods have achieved good results, accuracy is difficult to guarantee in complex conditions such as long detection distances and sandy weather. The performance of the models YOLOV5 and YOLOv7 on 7581 hard hat datasets was compared in this paper. As a result, YOLOv7 has the best detection performance on the helmet datasets, with 96.5% accuracy and 62FPS speed. This result represents that YOLOv7 has high effectiveness in helmet detection.