Safety Helmet Detection Based On YOLOV3N

Abstract

The YOLOv3 algorithm is widely used in the industry due to its high speed and high precision. Aiming at the problem of low detection accuracy and slow detection rate of wearing helmets in intelligent monitoring, a detection algorithm YOLOv3N based on improved YOLOv3 (You Only Look Once) is proposed. Improve the network structure on the basis of the YOLOv3 algorithm, replace the Darknet-53 traditional convolution with a convolution structure with fewer parameters, reduce model parameters, and increase the detection rate; in order to screen out the required detection frames more reasonably, the NMS is optimized. Experimental results show that compared with YOLOv3, YOLOv3N improves the number of frames per second (FPS) by 64%, and achieves an accuracy of 93.8%.