Chinese Journal of Liquid Crystals and Displays, Volume. 37, Issue 9, 1228(2022)
Real-time detection model of highway vehicle based on YOLOv5s
Fig. 2. Convolutional attention residual module.（a） ResBottleNeck； （b） CAM； （c） SAM.
Fig. 4. Convolutional attention pyramid network.（a） Backbone； （b） Up-bottom path augmentation； （c） Bottom-up path augmentaion.
Table 1. Datasets content description
View tableView in ArticleTable 1. Datasets content description
种类 描述 数量 Car 私家轿车、4座及以内车型 3 576 Van 小货车、商务车、8座及以内车型 2 398 Bus 客车 3 074 Truck 大型卡车、大型货车 3 685
Table 2. Comparison of detection accuracy of different algorithms on Ningxia expressway vehicle datasets
View tableView in ArticleTable 2. Comparison of detection accuracy of different algorithms on Ningxia expressway vehicle datasets
Detector Car Van Bus Truck mAP SSD 0.836 0.635 0.792 0.865 0.782 YOLOv3 0.906 0.658 0.889 0.927 0.845 YOLOv4 0.951 0.714 0.914 0.964 0.886 YOLOv5s 0.937 0.681 0.911 0.955 0.871 YOLOv5s-CRCP（ours） 0.986 0.744 0.923 0.995 0.912
Table 3. Comparison of test results of different algorithms on Ningxia expressway vehicle datasets
View tableView in ArticleTable 3. Comparison of test results of different algorithms on Ningxia expressway vehicle datasets
Detector Base Network FPS SSD VGG16 26 YOLOv3 DarkNet53 36 YOLOV4 CSPDarkNet53 38 YOLOv5s —— 77 YOLOv5s-CRCP（Ours） —— 75
Table 4. Ablation experiment of CR
View tableView in ArticleTable 4. Ablation experiment of CR
Index 实验1 实验2 实验3 实验4 mAP 0.871 0.887 0.876 0.893 FPS 77 76 77 76
Table 5. Ablation experiment of CP
View tableView in ArticleTable 5. Ablation experiment of CP
Index 实验1 实验2 实验3 实验4 mAP 0.871 0.894 0.887 0.901 FPS 77 76 76 75
Table 6. Final algorithm ablation experiment
View tableView in ArticleTable 6. Final algorithm ablation experiment
Index 实验1 实验2 实验3 实验4 mAP 0.871 0.893 0.901 0.912 FPS 77 76 75 75
Get Citation
Copy Citation Text
Yuan-feng LIU, Hai-jun JI, Li-bo LIU. Real-time detection model of highway vehicle based on YOLOv5s[J]. Chinese Journal of Liquid Crystals and Displays, 2022, 37(9): 1228
Category: Research Articles
Received: Jan. 24, 2022
Accepted: --
Published Online: Sep. 19, 2022
The Author Email: Li-bo LIU (liulib@163.com)
