Laser & Optoelectronics Progress, Volume. 62, Issue 12, 1237008(2025)
Improved Lightweight Traffic Sign Detection Algorithm for YOLOv8n
Figures & Tables(20)
Fig. 16. Detection results of improved YOLOv8n and original YOLOv8n algorithm. (a) Improved YOLOv8n; (b) original YOLOv8n
Table 1. Performance comparison of different algorithms
View tableTable 1. Performance comparison of different algorithms
Algorithm Precision /% Recall /% mAP@0.50 /% mAP@0.50∶0.95 /% FPS /（frame/s） Parameters /106 SSD 60.2 52.7 54.8 39.0 29.1 26.28 Faster R-CNN 58.2 50.7 52.8 38.6 22.0 28.34 YOLOv3 62.2 59.7 60.3 43.1 36.6 61.73 YOLOv4 63.7 61.2 63.7 44.9 39.4 64.36 YOLOv5s 72.9 70.7 70.1 52.0 47.0 7.02 YOLOv7-tiny 56.5 47.1 46.4 31.9 41.3 4.82 YOLOv8n 64.6 62.6 67.0 49.1 59.2 3.01 YOLOv5s-Ghost［ 22 ］65.0 54.0 59.3 44.1 34.4 5.00 Improved YOLOv8n［ 23 ］70.0 67.9 71.8 53.9 46.2 4.37 Ghost-YOLOv8［ 22 ］71.7 66.7 71.9 54.7 51.1 2.80 Proposed 77.7 64.7 74.6 55.3 47.8 2.78
Table 2. Results in ablation experiments
View tableTable 2. Results in ablation experiments
Baseline ODC_C2f EMA YOLOHead WIoU v3 P /% R /% mAP@0.50 /% mAP@0.50∶0.95 /% YOLOv8n 64.6 62.6 67.0 49.1 √ 65.7 63.1 67.7 49.9 √ 66.2 63.5 68.0 50.0 √ 75.0 67.1 73.1 54.6 √ 67.5 63.5 67.9 50.1 √ √ 67.6 63.7 68.4 50.3 √ √ √ 76.4 64.9 74.7 55.8 √ √ √ √ 76.6 67.8 76.2 56.5
Table 3. Lightweight improved ablation experiments
View tableTable 3. Lightweight improved ablation experiments
Baseline GhostNetV2_C2f GSConv Precision /% Recall /% mAP@0.50 /% Parameters /106 FLOPs /Gbit Model size /106 YOLOv8n+ODC_C2f+EMA+YOLOHead+WIoU v3 76.6 67.8 76.2 3.34 12.0 7.2 √ 74.1 65.3 73.8 2.88 11.8 6.3 √ 78.1 72.0 76.4 3.24 13.0 7.0 √ √ 77.7 64.7 74.6 2.78 11.7 6.1
Table 4. Performance comparison of different algorithms
View tableTable 4. Performance comparison of different algorithms
Algorithm P /% R /% mAP@0.50 /% FPS /（frame/s） Parameters /106 FLOPs /Gbit Model size/106 SSD 79.0 62.4 70.5 30 26.28 62.65 91.2 YOLOv3-tiny 82.4 68.3 76.3 79 8.67 12.90 17.4 YOLOv4-tiny 83.0 69.7 77.6 83 5.88 6.96 22.4 YOLOv5s 84.4 72.7 79.4 58 7.02 15.80 14.7 YOLOv7-tiny 80.2 66.2 73.2 51 4.82 10.30 10.0 YOLOv8n 85.4 71.8 78.1 74 3.01 8.20 6.3 Improved YOLOv5s［ 11 ］85.8 68.4 75.5 62 5.85 12.40 11.5 SM-YOLOv8［ 24 ］86.0 66.3 74.7 ‒ 2.11 ‒ ‒ Proposed 88.2 72.8 80.5 50 2.98 10.10 5.8
Tools
Get Citation
Copy Citation Text
Xianglong Luo, Wenxin Lü, Zhenyue Shi, Ruochen Liu. Improved Lightweight Traffic Sign Detection Algorithm for YOLOv8n[J]. Laser & Optoelectronics Progress, 2025, 62(12): 1237008
Paper Information
Category: Digital Image Processing
Received: Oct. 15, 2024
Accepted: Dec. 25, 2024
Published Online: Jun. 12, 2025
The Author Email: Xianglong Luo (xlluo@chd.edu.cn)
CSTR:32186.14.LOP242109
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20