Class Overwhelms: Mutual Conditional Blended-Target Domain Adaptation

3 Feb 2023  ·  Pengcheng Xu, Boyu Wang, Charles Ling ·

Current methods of blended targets domain adaptation (BTDA) usually infer or consider domain label information but underemphasize hybrid categorical feature structures of targets, which yields limited performance, especially under the label distribution shift. We demonstrate that domain labels are not directly necessary for BTDA if categorical distributions of various domains are sufficiently aligned even facing the imbalance of domains and the label distribution shift of classes. However, we observe that the cluster assumption in BTDA does not comprehensively hold. The hybrid categorical feature space hinders the modeling of categorical distributions and the generation of reliable pseudo labels for categorical alignment. To address these, we propose a categorical domain discriminator guided by uncertainty to explicitly model and directly align categorical distributions $P(Z|Y)$. Simultaneously, we utilize the low-level features to augment the single source features with diverse target styles to rectify the biased classifier $P(Y|Z)$ among diverse targets. Such a mutual conditional alignment of $P(Z|Y)$ and $P(Y|Z)$ forms a mutual reinforced mechanism. Our approach outperforms the state-of-the-art in BTDA even compared with methods utilizing domain labels, especially under the label distribution shift, and in single target DA on DomainNet. Source codes are available at \url{https://github.com/Pengchengpcx/Class-overwhelms-Mutual-Conditional-Blended-Target-Domain-Adaptation}.

PDF Abstract

Datasets


Introduced in the Paper:

Office-Home-LMT

Used in the Paper:

Office-Home DomainNet Office-31
Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Multi-target Domain Adaptation DomainNet MCDA Accuracy 34.5 # 1
Blended-target Domain Adaptation DomainNet MCDA Average Accuracy 34.5 # 1
Domain Adaptation DomainNet MCDA Average Accuracy 35.2 # 1
Blended-target Domain Adaptation Office-31 MCDA Average Accuracy 89.6 # 1
Multi-target Domain Adaptation Office-31 MCDA Accuracy 89.6 # 1
Blended-target Domain Adaptation Office-Home MCDA Average Accuracy 71.1 # 1
Multi-target Domain Adaptation Office-Home MCDA Accuracy 71.1 # 1
Label shift of blended-target domain adaptation Office-Home-LMT MCDA Average Accuracy 65.9 # 1

Methods