Vehicle Multi-attribute Recognition Based on Improved YOLOv5
Multi-attribute recognition of vehicles can be used for traffic monitoring and smart bayonet,which is of great signif-icance to traffic management.In order to improve the accuracy and reliability of vehicle detection and positioning in traffic monitor-ing scenarios,an attention mechanism is proposed based on the YOLOv5 algorithm.The modular approach enhances the ability of the network to extract features,and trains the three attributes of the vehicle's brand,color and car model,taking full account of the correlation between the attributes.The dataset uses the Cars Dataset publicly available on the Internet,and the dataset is expanded and preprocessed.Experimental results show that based on the improved YOLOv5 model,the recognition accuracy reaches 95.71%,which is 2.61%higher than the detection accuracy of the original YOLOv5 model.While ensuring the detection speed,it has higher recognition accuracy and can meet the requirements of video real-time.
deep learningYOLOv5vehicle detectionmulti-attribute recognition
国家科技期刊平台
NSTL
万方数据
