Detection of Surface Defects of Magnetic Tiles Based on Improved YOLOv5

The typical defect detection algorithm is ineffective due to the contrast between magnetic tile and various features. An improved YOLOv5-based algorithm, for detecting defects with varying features, suggested. procedure begins by incorporating CBAM into feature extraction network of YOLOv5. It improves learning capabilities target region filtering weighting vectors in such a way that processing dominated essential characteristics. A new loss function model then proposed according properties picture, confidence prediction box increased. Data augmentation technologies are introduced increase number data samples. Based on datasets, evaluation results have shown precision approach 98.56%, 3.21%, 7.22% greater than original YOLOv5 Faster R-CNN, respectively, all which demonstrate effectiveness accuracy method.