Research on Helmet Wearing Detection in Multiple Scenarios Based on YOLOv5

Wearing a helmet can effectively protect the human head in industrial production and traffic activities. In order to monitor whether relevant people are wearing helmets in real time, YOLOv5 target detection algorithm is combined with helmet wearing detection in this paper, and the head and helmet dataset are pre-trained through its own yolov51.pt weight file to obtain the characteristics of the head and helmet. Then the test pictures of different scenarios are sent to the network for performance verification of the algorithm, and compared with YOLOv4 in the quantitative analysis. Finally, the score of mAP equals to 0.924 is obtained. The results show that the reference value for auxiliary real-time detection of helmet wearing can be provided by this algorithm.