3,570 images of metal surfaces categorized into 10 distinct defect classes such as punching, welding line, and crescent gap. These images include bounding box annotations for localizing surface irregularities in industrial manufacturing environments.
Use Cases
- Train object detection models to identify and localize 'punching' or 'welding line' defects using the provided bounding box coordinates.
- Develop multi-class classification algorithms to distinguish between 'water spot' and 'oil spot' anomalies.
- Benchmark the performance of anomaly detection systems on 'inclusion' and 'rolled pit' defects in industrial textures.
Strengths
- 3,570 images of metal surfaces with real-world industrial defects.
- 10 defect categories including punching, welding line, crescent gap, water spot, oil spot, silk spot, inclusion, rolled pit, crease, and waist fold.
- Bounding box coordinates provided for each defect instance to facilitate object detection training.