Steel Surface Defect Detection Algorithm Based on Improved YOLOv5
In view of the problem of insufficient detection accuracy due to size difference between different defects during steel surface defect detection,this paper proposes the detection algorithm based on improved YOLOv5.By introducing CA attention mechanism module into the network structure,this algorithm helps to enhance the focusing ability of the model on valid sample information and restrains interference of irrelevant factors.In addition,the algorithm integrates context convolution module on the basis of Spatial Pyramid Pooling(SPP)module to enhance feature representation.Data set verification results show that the present model is excellent at identifying surface defects of different types of steel,with average precision average(mAP)increased by 5.2%compared with the original YOLOv5 algorithm,it can meet the practical needs of steel defect detection on the production site.
NETL
NSTL
万方数据
