CoI2A: Collaborative Inter-domain and Intra-domain Alignments for Multisource Domain Adaptation

In the remote sensing information interpretation tasks, compared with collecting lots of high-quality image labels for the target domain, a large amount of labeled remote sensing data from multiple source domains are generally available without any extra cost. In this article, our work focuses on how to exploit the rich knowledge obtained from multiple source domains to guide the interpretation of the target scene, and we propose a novel framework called collaborative interdomain and intradomain alignments for multisource domain adaptation (MDA), namely CoI
$^{2}\text{A}$
, in which interdomain and intradomain alignments are well collaborated to reduce the distribution divergence across sources and target. To reduce the discrepancy across sources, the intersource alignment is proposed to map multiple sources into a unified representation space. In addition, the cross-domain attention is introduced to enforce the intraclass compactness of the target. Interdomain alignment aligns each source with target domain separately with the help of cross-domain attention. As for the intradomain alignment, the multihead attentive representations of the target obtained by cross-domain attention are correlated into a unified one. The experimental results obtained from different scene classification tasks demonstrate the superiority of our model.