Table 5 Comparison of detection accuracy of models on different defects.

From: Research on steel surface defect classification method based on deep learning

Model

crazing

inclusion

patches

pitted

_surface

rolled-in_scale

scratches

mAP

YOLOv5l

0.537

0.836

0.895

0.734

0.668

0.874

0.757

YOLOv5-KBS

0.580

0.828

0.922

0.841

0.652

0.970

0.799

  1. Significant values are in bold.
Back to article page