Balanced multi-scale target score network for ceramic tile surface defect detection

Surface defect detection of ceramic disc based on improved YOLOv5s

“…Currently, the majority of research on ceramic tile defect detection focuses on deep learning-based visual detection methods, while there is comparatively less research on daily ceramic defect detection. The balanced multiscale target scoring network algorithm, proposed by Cao et al for ceramic tile surface defect detection, enhances the YOLOv5s algorithm by introducing content-aware feature recombination and dynamic attention mechanisms [ 5 ].The proposed method achieved an average precision improvement of 4.9, with a 6 % increase in AP for small targets. A supervised automatic detection method for ceramic tile surface defects, proposed by Lu et al, introduces depth separable convolution, incorporates DWBottleneck and residual connections, and replaces the feature extraction backbone of YOLOv5s with an improved Shufflenetv2 [ 6 ].…”