“Remote sensing image semantic segmentation plays an important role in land cover and use classification, urban planning, and change detection. As a highly promising unsupervised learning method, domain adaptation technology has greatly promoted the development of semantic segmentation in remote sensing images. However, existing models are still based on single task learning, and the segmentation features obtained from learning are not sufficient, resulting in difficulty in accurately identifying complex regions in remote sensing images during the segmentation process. To address this issue, experts have proposed a multi task learning domain adaptive semantic segmentation network MTLDANet, which enhances the learning ability of segmentation features by synergistically learning semantic and elevation information in remote sensing images.”