A Vehicle Detection Method Based on Improved YOLOV5s

Abstract

Along with the advancement of artificial intelligence technology, the development of autonomous driving has been expanding. Vehicle detection is an indispensable part of autonomous driving technology. To achieve high accuracy and high speed, this paper proposes a vehicle detection method based on improved YOLOV5s. Firstly, the Convolutional Block Attention Module (CBAM) is introduced, which enhances the ability of the network to extract vehicle features and improves the detection accuracy significantly. Secondly, some ordinary convolutions are replaced by the Ghost Module, which is a lightweight convolutional module in Backbone in order to reduce the computational cost and the number of parameters significantly. Finally, experimental results show that the proposed method improves the accuracy by 2 % over the original YOLOV5s, reduces the parameter quantity to 82.97% of the original model, and achieves the target detection of vehicles effectively.