Research on Lightweight Object Detection of YOLOv5 Based on Vehicle Vision in Urban Rail Transit
YOLO series algorithm has been widely studied because of its advantages,but its complex model restricts itsapplication in some scenarios.In view of the timeliness of foreign object intrusion detection in the front track area during the driving of unmanned trains in urban rail transit,this paper,based on YOLOv5s algorithm,optimizes the lightweight design of the backbone network and the regression loss function of the boundary box.First,the GhostNet network is introduced to reduce the network parameters and improve the lightweight of the model.Then,DIOU is used as the boundary box regression loss function to improve the convergence speed and the accuracy of the model.The experimental results on COCO data set show that the proposed algorithm improves the model lightness and detection accuracy compared with YOLOv5s.
NETL
NSTL
万方数据
