A Vehicle Detection Method Based on YOLOV4 Model
Haolong Peng, Song Guo, Xiaoyi Zuo · 2021
Vehicle detection is a basic problem in target detection and has significant application value in fields such as intelligent traffic supervision and automatic driving. In order to improve the vehicle detection accuracy and detection speed, a vehicle detection method based on YOLOv4 is used in this paper. First, manually label the Apollo Scape automatic driving data set to establish a vehicle detection database, and then the YOLOv4 deep learning model is used to extract vehicle features adaptively, which avoiding the artificial subjective selection of features and achieving high precision vehicle target detection. The experimental results show that this model has a significant improvement in detection accuracy. Specifically, the mean average precision (mAP) reaches 92.1%, which is higher than the YOLOv4Tiny, YOLOv3 and Faster RCNN models 3.2%, 6.9% and 3.1%, respectively.