GC10-DET — Metallic Surface Defect Dataset

Mirror pulled via the Roboflow CC BY 4.0 mirror (version 5, COCO export), not the original Baidu Pan link (awkward to access outside China, license unstated there). Staged as a bbox-supervised detector warmup dataset for CoRe's Situational Control paint-inspection work (core-lab/situational-control).

Source

• Original repo (Baidu Pan download, license unstated): https://github.com/lvxiaoming2019/GC10-DET-Metallic-Surface-Defect-Datasets
• Roboflow mirror (source of this copy, CC BY 4.0 confirmed): https://universe.roboflow.com/nanjing-university-of-information-science-and-technology/gc10-det-zukta
• License: CC BY 4.0 (per Roboflow mirror)

Contents

2,294 grayscale line-scan images, 10 defect classes (punching, weld line, crescent gap, water/oil/silk spot, inclusion, rolled pit, crease, waist fold), bounding-box annotated, COCO format. Bare steel sheet, not painted — useful for bbox detector pretraining, not a domain match for the paint pilots.

Citation

Lv, X., Duan, F., Jiang, J.J., Fu, X., Gan, L. "Deep Metallic Surface Defect Detection: The New Benchmark and Detection Network." Sensors 2020, 20(6), 1562.