MemSAC: Memory Augmented Sample Consistency for Large Scale Unsupervised Domain Adaptation
This addresses the problem of small inter-class discriminability in large-scale unsupervised domain adaptation for practical real-world datasets, representing a strong specific gain rather than a foundational advancement.
The paper tackles the challenge of unsupervised domain adaptation with many categories by proposing MemSAC, which uses sample-level similarity and a novel contrastive loss to achieve discriminative transfer, resulting in significant improvements over state-of-the-art methods on large-scale datasets like DomainNet with 345 classes and Caltech-UCSD birds with 200 classes.
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.