47,420 road images collected from six countries including Japan, India, Czech Republic, China, Norway, and the United States, featuring 12,135 annotated damage instances. The dataset categorizes road distress into four classes: longitudinal cracks, transverse cracks, alligator cracks, and potholes.
Use Cases
- Train object detection models like YOLO or SSD using the bounding box coordinates and damage class labels
- Evaluate model robustness and transfer learning performance across different country-specific subsets
- Develop automated infrastructure maintenance systems that categorize road distress using the D00-D40 classification schema
Strengths
- Contains 47,420 images with bounding box annotations for 12,135 damage instances
- Standardized labels across four categories: D00 (Longitudinal Crack), D10 (Transverse Crack), D20 (Alligator Crack), and D40 (Pothole)
- Annotations provided in Pascal VOC XML format including xmin, ymin, xmax, and ymax coordinates
- Data sourced from diverse geographical regions including Asia, Europe, and North America