Welding seam defect detection method based on semi-supervised transfer learning

The invention relates to the field of welding seam quality detection, in particular to a welding seam defect detection method based on semi-supervised transfer learning, which comprises the following steps: acquiring an original image; creating a data set P1 and a data set P2; preprocessing the enhanced image in the data set P2; building a multi-domain learning network based on semi-supervised transfer learning; inputting the enhanced image which is not marked with the welding seam feature in the data set P1 into a backbone network to train the backbone network, obtaining a secondary enhanced image which is output by the backbone network and is provided with an image welding seam feature prediction mark, and inputting network parameters of the trained backbone network into a secondary network; training a secondary network by using the preprocessed data set P2 and a secondary enhanced image; inputting a detection image containing the quality of a to-be-detected welding seam into the trained multi-domain learni