Unsupervised Domain Adaptation Semantic Segmentation for Remote-Sensing Images via Covariance Attention
Yikun Liu, Xudong Kang, Yuwen Huang, Kuikui Wang, Gongping Yang
Abstract
Semantic segmentation for remote sensing is a crucial but challenging task. Many supervised semantic segmentation methods rely heavily on a large-scale pixel-wise annotated data set, but it is time-consuming and laborious to provide manual annotation. However, due to the common domain shift of remote sensing images, a direct transfer might not perform well. Therefore, many unsupervised domain adaptation methods have been proposed to solve the data distribution discrepancy in remote-sensing data sets, but these methods cannot completely utilize the features extracted in the training process. In addition, the correlations between feature map channels are crucial for the pixel-wise classification task. In this letter, a covariance-based channel attention module is proposed to capture correlations by covariance metric and weighting the feature map channels. To further improve the domain adaptation performance, we propose a three-stage unsupervised domain adaptation semantic segmentation method for remote-sensing images, we fine-tune the model which has been trained on the source domain on the target domain via self training and knowledge distillation. To test the effectiveness of the proposed method, experiments are conducted on the ISPRS 2-D Semantic Labeling data set and an urban drone data set. Our method shows a better performance advantage compared with other state-of-the-art methods.