Opto-Electronic Engineering, Volume. 51, Issue 6, 240055-1(2024)
A traffic sign recognition method based on improved YOLOv5
Figures & Tables(18)
Fig. 7. Removing the path EFPN' from the large object detection layer
Fig. 11. Comparison of accuracy effects of traffic signs before and after improvement
Fig. 12. Comparison of missed detection effects of traffic signs before and after improvement
Fig. 13. Comparison of false detection effects of traffic signs before and after improvement
Table 1. Comparison of EFPN and EFPN' structures
View tableView in ArticleTable 1. Comparison of EFPN and EFPN' structures
改进方法 精确率P/% Params/M pn w32 ph4 模型说明：YOLOv5+PC3+EFPN表示YOLOv5中C3中的普通卷积用轻量化部分卷积PConv替换，构成PC3替换掉原C3结构，检测头用EFPN替换；YOLOv5+PC3+EFPN'表示YOLOv5中C3中的普通卷积用轻量化部分卷积PConv替换，构成PC3替换掉原C3结构，检测头用EFPN'替换。 YOLOv5+PC3+EFPN 0.93 0.70 0.76 4.96 YOLOv5+PC3+ EFPN ' 0.96 0.96 0.98 3.80
Table 2. YOLOv5 default anchor box size
View tableView in ArticleTable 2. YOLOv5 default anchor box size
检测尺度 Anchor1 Anchor2 Anchor3 小尺寸 [10,13] [16,30] [33,23] 中尺寸 [30,61] [62,45] [59,119] 大尺寸 [116,90] [156,198] [373,326]
Table 3. Results of K-means clustering algorithm
View tableView in ArticleTable 3. Results of K-means clustering algorithm
检测尺度 Anchor1 Anchor2 Anchor3 小尺寸 [5,5] [6,7] [8,9] 中尺寸 [9,14] [9,14] [14,15] 大尺寸 [19,20] [19,20] [25,26]
Table 4. Results of ablation experiment
View tableView in ArticleTable 4. Results of ablation experiment
编号 模型 P R mAP0.5 FPS L/ms Params/M 模型说明：FOG代表扩充雾化数据集TT100K-FOG；PC3代表使用更加轻量的PConv构建PC3特征提取模块来取代YOLOv5骨干和颈部网络中的C3模块；EFPN代表采用延伸的特征金字塔结构，替代YOLOv5中检测头；EFPN'代表在EFPN结构中删除大目标检测层后，替代YOLOv5中检测头；Focal-EIoU代表采用Focal-EloU取代YOLOv5默认函数CIoU；CBAM代表在YOLOv5主干网络中嵌入空间和通道注意力模块。 0 YOLOv5 0.842 0.824 0.861 145.7 6.9 7.10 1 YOLOv5+FOG 0.893 0.838 0.870 145.7 6.9 7.10 2 YOLOv5+FOG+PC3 0.853 0.768 0.840 166.7 6.0 4.87 3 YOLOv5+FOG+PC3+EFPN 0.862 0.759 0.842 135.1 7.4 4.96 4 YOLOv5+FOG+PC3+EFPN' 0.876 0.782 0.854 161.3 6.2 3.80 5 YOLOv5+FOG+PC3+EFPN'+Focal-EIoU 0.906 0.790 0.860 161.3 6.2 3.80 6 YOLOv5+FOG+PC3+EFPN'+Focal-EIoU+CBAM 0.917 0.853 0.899 151.5 6.7 3.95
Table 5. Performance comparison with other algorithms
View tableView in ArticleTable 5. Performance comparison with other algorithms
模型 平台 主干网 类型 P/% mAP0.5 /% FPS Faster R-CNN MMDetection ResNet50 Anchor-based 71.9 79.9 57.7 YOLOv4 Darknet Darknet Anchor-based 58.7 82.2 80.9 YOLOv5 YOLOv5 Darknet Anchor-based 84.2 86.1 145.7 YOLOX MMDetection Darknet Anchor-free 72.6 79.7 93.6 YOLOv6 YOLOv6 EfficientRep Anchor-free 77.7 81.1 162.8 YOLOv7 YOLOv7 E-ELAN Anchor-based 72.0 77.4 130.2 YOLOv8 YOLOv8 Darknet Anchor-free 87.7 83.7 171.4 Ours YOLOv5 Darknet Anchor-based 91.7 89.9 151.5
Tools
Get Citation
Copy Citation Text
Liguo Qu, Xin Zhang, Zibao Lu, Yuling Liu, Guohao Chen. A traffic sign recognition method based on improved YOLOv5[J]. Opto-Electronic Engineering, 2024, 51(6): 240055-1
Paper Information
Category:
Received: Mar. 7, 2024
Accepted: Jun. 4, 2024
Published Online: Oct. 21, 2024
The Author Email: Qu Liguo (曲立国)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20