Lightweight YOLO v7 Fatigue Detection Method Based on BiFPN Fusion
To address the issue of low fatigue detection accuracy and poor real-time performance of hoist drivers,this paper proposes a lightweight fatigue detection method based on BiFPN fusion.This model chooses YOLO v7,which has a faster single-stage detection speed,as the target detector the redundant convolution calculations in the YOLO v7 backbone network are replaced with a lightweight Ghost network to extract features,and the exponential activation function in Ghost network is replaced by lightweight SMU(Smooth Maximum Unit)activation function.The lightweight YOLO v7 fatigue detection model with Bidirectional feature pyramid(BiPFN)is tested on the self-built dataset of mine hoist driver fatigue driving.The results show that the average accuracy reached 97.25%,the real-time performance reached 78 FPS(frames per second),which increased the accuracy by 3.14%and the speed by 8 FPS compared with the original YOLO v7 network.