Multi-source domain adaptive semantic segmentation method and device based on multi-level domain correlations

Abstract

Disclosed in the present invention is a multi-source domain adaptive semantic segmentation method based on multi-level domain correlations, including: (1) pre-training a cross-domain semantic segmentation model F on all source domains and a target domain; calculating multi-level domain correlations based on the pre-trained F, including domain-level source-target correlations dand pixel-level source-target correlations constructing source-target domain mixed images and corresponding pseudo-labels based on constructing source-target domain mixed images and corresponding pseudo-labelsbased on a random sampling method; training the F on all the source domains and the target domain based on d, and performing, by the trained F, multi-source domain adaptive semantic segmentation on an image to be detected to extract a target object in the image. This method can improve weights of the source domains and the pixels in the source domains with high correlations with the target domain, and reduce weights of the source domains and the pixels in the source domains with low correlations with the target domain, so as to avoid interference of uncorrelated information of multi-source domains on the training.