Table 2 Comparison of detection results between YOLO-LRDD and the other five methods
From: YOLO-LRDD: a lightweight method for road damage detection based on improved YOLOv5s
Model | mAP0.5 (%) | mAP0.95 (%) | Precision (%) | Recall (%) | F1-Score (%) | #Parama. (M) | FLOPs (G) | FPS |
|---|---|---|---|---|---|---|---|---|
YOLOv5s | 56.9 | 29.4 | 58.9 | 56.2 | 57.5 | 27.6 | 28.4 | 72 |
YOLOv5m | 57.3 | 38.9 | 58.5 | 57.3 | 57.8 | 33.2 | 32.2 | 64 |
MobileNetv3-YOLOv5s | 51.2 | 25.8 | 54.7 | 53.8 | 54.2 | 19.2 | 24.2 | 78 |
GhostNet-YOLOv5s | 49.5 | 24.5 | 53.6 | 52.5 | 53.0 | 20.4 | 23.4 | 77 |
Faster R-CNN | 60.2 | 31.8 | 62.7 | 59.3 | 60.1 | 38.3 | 34.6 | 51 |
YOLO-LRDD | 57.6 | 30.6 | 59.2 | 58.2 | 58.7 | 19.8 | 17.4 | 86 |