Research on Fast Insulator Positioning Algorithm of Lightweight FAST-YOLOV5N

In this paper, a FAST-YOLOV5N Fast location algorithm for lightweight insulators is designed based on the network model of YOLOv5n. This algorithm replaces the original backbone network module with the more efficient ShuffleNetv2 basic unit, which effectively improves the computational efficiency of the algorithm. Secondly, in the feature fusion part of the network, we use depthwise separable convolution and ReLU activation functions, which help to further improve the inference speed of the network. Finally, in the training strategy, we use the SIoU loss function to accelerate the convergence speed and positioning accuracy of the network. Through experimental comparison, compared with the original YOLOv5n model, the proposed model achieves 50% speed improvement with only 0.8% mAP accuracy loss, which effectively improves the speed of the model algorithm.