A high-effective multitask surface defect detection method based on CBAM and atrous convolution

Given the shortcomings of conventional machine vision-based surface defect detection methods, including their low accuracy, long development cycle, and poor generalization ability, this paper proposes a surface defect detection model based on the convolutional block attention module and atrous convolution. This model combines the surface defect segmentation task of the product with the classification task, obtains contextual information of the image at multiple scales using atrous spatial pyramid pooling, and then uses the convolutional block attention module to reallocate the weighting of the network to enhance focus on the defect area and improve the discrimination of extracted features. In addition, atrous convolution was introduced in the deep network to simplify the model when used in defect segmentation tasks and enhances the real-time performance of the model defect detection method. Experiments show the superior accuracy and real-time performance of the proposed model when compared with current mainstream surface defect detection methods and indicate its wide applicability in the detection of surface defects in industrial products.