SABIQ — Road Damage Detection Model
Proactive road defect detection system.
Model Details
- Architecture: YOLO26m
- Base Model: yolo26m.pt (Ultralytics)
- Dataset: RDD2022 (Road Damage Detection 2022)
- Classes: crack, other, pothole
- mAP50: 0.636
- Epochs: 65
- Image Size: 640
- Training Hardware: NVIDIA A100
Validation Results
| Class |
Images |
Instances |
Precision |
Recall |
mAP50 |
mAP50-95 |
| all |
5758 |
9737 |
0.687 |
0.585 |
0.636 |
0.349 |
| crack |
3266 |
7209 |
0.714 |
0.520 |
0.605 |
0.321 |
| other |
1093 |
1563 |
0.714 |
0.745 |
0.792 |
0.493 |
| pothole |
544 |
965 |
0.635 |
0.491 |
0.512 |
0.233 |
Classes
| ID |
Label |
Description |
| 0 |
crack |
Longitudinal, transverse and alligator cracks |
| 1 |
other |
Other road corruption |
| 2 |
pothole |
Road potholes |
Live API
https://rahaf2001-sabiq-api.hf.space
