Aiming at the low detection accuracy of the current algorithm in practical application ofstrip surface defect detection

a method of strip surface defect detectionbased on the improved YOLOv5s method is proposed. Based on YOLOv5s model

firstly

the Swin Transformer structure is fused in the backbone network

and the feature is fused with the neck network to enhance the feature extraction ability of strip steel surface defects. Secondly

the global attention mechanism is integrated into the C3 structure to enlarge the global cross-latitude interactive features and improve the detection efficiency while reducing the dispersion of feature information. Finally

the decoupling head is used to replace the detection head in the model to better solve the contradiction between the classification task and the regression task. The improved method of strip surface defect detection is tested via data set. The results show that the precision of the improved method reaches 85.0%

and the mean average precision reaches 80.8%

which is 9.5% and 5.7% higher than the original YOLOv5s algorithm

further meeting the requirements for the detection accuracy of strip surface defects.