metadata
license: unknown
RoadEye Unified Pothole Detection Dataset for YOLO
The RoadEye dataset aggregates annotated and augmented road-scene images from Roboflow, Mendeley, Kaggle, and custom synthetic collections for pothole detection and road-condition analysis. It provides a consistent YOLO annotation format across ~12 k images, split into training, validation, and test sets.
To access training scripts go here: https://github.com/parthubhe/RoadEye_Backend
- Domain: Autonomous driving / road monitoring
- Tasks: Object detection, instance segmentation
- Format: YOLO (bounding box annotations)
- Size: ~12 k images, ~25 k annotated instances
- License: Combination of Roboflow, Mendeley, Kaggle, and CC-BY-4.0 (see individual source licenses)