Research Article
YOLOv5-PD: A Model for Common Asphalt Pavement Defects Detection
Table 1
Experiment results with different improvement methods.
| | mAP % | MeshCracks AP% | Potholes AP % | Pacthes AP % | FPS |
| | YOLOv5l | 68.5 (0) | 69.7 (0) | 65.3 (0) | 70.4 (0) | 105 | | YOLOv5l + CBAM | 71.2 (+2.7) | 69 (-0.7) | 68.9 (+3.6) | 75.7 (+5.3) | 102 | | YOLOv5l + BigKernel | 72.8 (+4.3) | 74.1 (+4.4) | 65.1 (-0.2) | 79.4 (+9.0) | 48 | | YOLOv5l + BigKernel + CBAM | 73.3 (+4.8) | 72.1 (+2.4) | 67.7 (+2.4) | 80.1 (+9.7) | 41 |
|
|
