Safety Equipment Wearing Detection Algorithm for Electric Power Workers Based on RepGFPN-YOLOv5

Abstract Wearing inspection safety equipment such as insulating gloves and safety helmets is an important guarantee for safe power operations. Given the low accuracy of the traditional insulating gloves and helmet-wearing detection algorithm and the problems of missed detection and false detection, this paper proposes an improved safety equipment wearing detection model named RepGFPN-YOLOv5 based on YOLOv5. This paper first uses the K-Means + + algorithm to analyze the data set for Anchor parameter size re-clustering to optimize the target anchor box size; secondly, it uses the neck network (Efficient Reparameterized Generalized Feature Pyramid Network, RepGFPN), which combines the efficient layer aggregation network ELAN and the re-parameterization mechanism), to reconstruct the YOLOv5 neck network to improve the feature fusion ability of the neck network; reintroduce the coordinate attention mechanism (Coordinate Attention, CA) to focus on small target feature information; finally, use WIoU_Loss as the loss function of the improved model to reduce prediction errors. Experimental results show that the RepGFPN-YOLOv5 model achieves an accuracy increase of 2.1% and an mAP value of 2.3% compared with the original YOLOv5 network, and detection speed of the improved model reaches 89FPS.The code: https://github.com/CVChenXC/RepGFPN-YOLOv5.git.