Dynamic Transfer for Multi-Source Domain Adaptation

Yinpeng Chen2

Lu Yuan2


UC San Diego1, Microsoft2

Overview


Recent works of multi-source domain adaptation focus on learning a domain-agnostic model, of which the parameters are static. However, such a static model is difficult to handle conflicts across multiple domains, and suffers from a performance degradation in both source domains and target domain. In this project, we present dynamic transfer to address domain conflicts, where the model parameters are adapted to samples. The key insight is that adapting model across domains is achieved via adapting model across samples. Thus, it breaks down source domain barriers and turns multi-source domains into a single-source domain. This also simplifies the alignment between source and target domains, as it only requires the target domain to be aligned with any part of the union of source domains. Furthermore, we find dynamic transfer can be simply modeled by aggregating residual matrices and a static convolution matrix. Experimental results show that, without using domain labels, our dynamic transfer outperforms the state-of-the-art method by more than 3% on the large multi-source domain adaptation datasets -- DomainNet.

paper

Published in IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2021.

Arxiv

Repository

Bibtex

Models


paper

Architecture: Dynamic residual transfer on ResNet bottleneck block.

Highlights

Replace domain alignment by sample alignment with input dependent dynamic network.

Benefit

  1. Better adaptation performance.
  2. The model does not rely on domain labels among source domains.

Results


paper

Results of DRT: The adaptation performance of DRT on DomainNet dataset.

Video


Authors



Yunsheng Li

UC San Diego