Lightweight Traffic Sign Recognition and Detection Algorithm Based on Improved YOLOv5s

Fig. 11. Detection results of BMGE-YOLOv5s on CCTSDB2017 and TSRD datasets. (a) Low illumination or at night; (b) high illumination; (c) complex background information; (d) signs partially worn or obscured; (e) multi-scale, multi-perspective; (f) normal illumination

Download full size

Table 1. Experimental environment configuration
View table
Table 1. Experimental environment configuration
Parameter Configuration
CPU Intel Core i7-13700KF
GPU NVIDIA RTX A5000 GDDR6
Video storage 24 G
Memory 128 G
Operating system Ubuntu 20.04
Graphics card acceleration CUDA 11.8
Programming language Python 3.9

Table 2. Performance comparison of different algorithms

View table

Table 2. Performance comparison of different algorithms

Algorithm	Parameters / $10^{6}$	FLOPs / $10^{9}$	Weight /M	FPS /（frame·s^-1）	mAP@0.5 /%
YOLOv4	64.40	142.8	245.0	67	90.04
YOLOv5s	7.03	16.0	14.2	181	89.80
YOLOv7	36.90	104.7	156.0	126	92.70
YOLOv8	11.72	28.5	5.9	159	94.53
DETR	72.31	47.6	252.0	94	91.67
BMGE-YOLOv5s	5.20	14.5	14.0	303	93.10

Table 3. Performance comparison of different loss functions
View table
Table 3. Performance comparison of different loss functions
Model P R mAP@0.5 mAP
YOLOv5s+IoU 84.9 86.1 87.8 84.3
YOLOv5s+DIoU 86.3 88.3 88.3 84.9
YOLOv5s+CIoU 90.5 88.1 89.8 85.1
YOLOv5s+MPDIoU 89.8 87.0 90.0 85.2

Table 4. Results of ablation experiments
View table
Table 4. Results of ablation experiments
Model P /% R /% mAP@0.5 /% FPS /（frame·s^-1）
YOLOv5s 85.5 88.1 89.8 181
YOLOv5s+BoTNet 81.8 95.1 91.9 185
YOLOv5s+MPDIoU 89.8 87.0 90.0 188
YOLOv5s+BoTNet+MPDIoU 85.7 89.6 92.3 196

Table 5. Results of lightweight ablation study
View table
Table 5. Results of lightweight ablation study
Model Number of layers Weight /M Parameters /10⁶ FLOPs /10⁹ FPS /（frame·s^-1） mAP@0.5 /%
YOLOv5s 270 14.2 7.03 16.0 359 93.40
YOLOv5s+GBneckv2 274 13.6 7.16 16.1 357 93.70
YOLOv5s+GBneckv2+ECA 284 12.4 6.51 15.5 345 94.10

Tools

Get Citation

Copy Citation Text

Fei Liu, Yanfen Zhong, Jiawei Qiu. Lightweight Traffic Sign Recognition and Detection Algorithm Based on Improved YOLOv5s[J]. Laser & Optoelectronics Progress, 2024, 61(24): 2412004

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites