Safety Helmet Monitoring System Based on Improved YOLOv5

In construction site, it is necessary to ensure that everyone wears helmet. Using object detection methods to do the supervising work is much more efficient than any other ways. First, this paper proposes a new helmet detection method based on improved YOLOv5 that can get better detection results in construction site. Second, this paper designs an automatically tracking and monitoring system to implement the real-time detection of the input. Third, this paper designs a UI interface to show all the details of the detection process. The monitoring system mainly includes the image processing module and the helmet detection module which are based on PyTorch framework and OpenCV image processing library. The dataset that we used for training and validation is SHWD.