A feature-based transfer learning method for surface defect detection in smart manufacturing

The employment of deep learning architecture for defect detection in the manufac-turing industry has gained due attention owing to the advancement of computa-tional technology. Conventional means of defect detection by manual visual in-spection by operators are often deemed laborious as well as prone to mistakes. In the present study, a feature-based transfer learning approach is used to classify surface defects. The KolektorSDD database is used in the present study. Two pipelines were developed to investigate its efficacy in detecting the defects, name-ly the VGG16-kNN and VGG16-SVM pipelines, respectively. It was demon-strated from the study that the VGG16-SVM pipeline was more superior com-pared to the VGG16-kNN pipeline as no misclassification transpired in either the test or the validation dataset. It could be concluded that the proposed pipeline is suitable for the classification of surface defects.