Helmet wearing detection algorithm based on improved YOLOv5

In industrial production, workers need to wear safety helmets at all times. However, due to different lighting, viewing angles, and the tendency of people to block each other, the precision of target detection is not high enough. Aiming at this problem, a real-time detection of helmets was achieved by improving the YOLOv5 algorithm. This algorithm introduces the lightweight network structure FasterNet, which uses partial convolution as the main operator to reduce the amount of calculations and parameters of the network; the boundary regression loss function Wise-IoU loss function with a dynamic focusing mechanism replaces the original loss function in YOLOv5; finally, the CBAM attention mechanism is introduced to obtain global context information and improve the detection ability of small targets. The experimental results show that the parameters of the improved YOLOv5 model are reduced by 12.68%, the computational amount is reduced by 10.8%, the mAP is increased from 88.3 to 92.3%, and the inference time is reduced by 81.5%, which is better than the performance of the original model and can detect helmet wearing effectively and in real time.