Research on Steel Surface Defect Detection Algorithm Based on Depth Learning
In order to solve the problem of insufficient intelligence and lightweight degree of steel surface defect detection model in the industrial field,so as to improve the adaptability and wide application of industrial machines,a lightweight steel surface defect detection model is proposed.Replace the main feature extraction network of YOLOv4 with MobileNetV3.At the same time,the ordinary convolution in the PANet network structure in YOLOv4 is replaced by the deeply separable convolution to improve the feature extraction capability;then,CA attention mechanism is introduced to enhance the ability of network feature perception;finally,the pruning technology is used to further compress the model,so as to build the YOLOv4 MEL model.The experimental results show that the optimized model performs better on the NEU-DET steel surface defect data set published by Northeastern University,with the mAP reaching 95.3%,and the amount of model parameters is small.Compared with YOLOv4 model,this method can effectively achieve steel surface defect detection.
万方数据
维普
