MemSAC: Memory Augmented Sample Consistency for Large Scale Unsupervised Domain Adaptation

TL;DR

MemSAC introduces a memory-augmented contrastive learning approach for large-scale unsupervised domain adaptation, effectively handling numerous classes and improving discriminative transfer between source and target domains.

Contribution

The paper proposes a scalable memory-augmented method with a novel contrastive loss for discriminative transfer in large-scale unsupervised domain adaptation.

Findings

Significant improvements on DomainNet with 345 classes

Effective adaptation on Caltech-UCSD birds dataset with 200 classes

Enhanced discriminative transfer through memory-augmented contrastive learning

Abstract

Practical real world datasets with plentiful categories introduce new challenges for unsupervised domain adaptation like small inter-class discriminability, that existing approaches relying on domain invariance alone cannot handle sufficiently well. In this work we propose MemSAC, which exploits sample level similarity across source and target domains to achieve discriminative transfer, along with architectures that scale to a large number of categories. For this purpose, we first introduce a memory augmented approach to efficiently extract pairwise similarity relations between labeled source and unlabeled target domain instances, suited to handle an arbitrary number of classes. Next, we propose and theoretically justify a novel variant of the contrastive loss to promote local consistency among within-class cross domain samples while enforcing separation between classes, thus preserving discriminative transfer from source to target. We validate the advantages of MemSAC with significant improvements over previous state-of-the-art on multiple challenging transfer tasks designed for large-scale adaptation, such as DomainNet with 345 classes and fine-grained adaptation on Caltech-UCSD birds dataset with 200 classes. We also provide in-depth analysis and insights into the effectiveness of MemSAC.