Abstract:Test-time domain adaptation aims to adapt the model trained on source domains to unseen target domains using a few unlabeled images. Emerging research has shown that the label and domain information is separately embedded in the weight matrix and batch normalization (BN) layer. Previous works normally update the whole network naively without explicitly decoupling the knowledge between label and domain. As a result, it leads to knowledge interference and defective distribution adaptation. In this work, we propose to reduce such learning interference and elevate the domain knowledge learning by only manipulating the BN layer. However, the normalization step in BN is intrinsically unstable when the statistics are re-estimated from a few samples. We find that ambiguities can be greatly reduced when only updating the two affine parameters in BN while keeping the source domain statistics. To further enhance the domain knowledge extraction from unlabeled data, we construct an auxiliary branch with label-independent self-supervised learning (SSL) to provide supervision. Moreover, we propose a bi-level optimization based on meta-learning to enforce the alignment of two learning objectives of auxiliary and main branches. The goal is to use the auxiliary branch to adapt the domain and benefit main task for subsequent inference. Our method keeps the same computational cost at inference as the auxiliary branch can be thoroughly discarded after adaptation. Extensive experiments show that our method outperforms the prior works on five WILDS real-world domain shift datasets. Our method can also be integrated with methods with label-dependent optimization to further push the performance boundary. Our code is available at <a class="link-external link-https" href="https://github.com/ynanwu/MABN" rel="external noopener nofollow">this https URL</a>.

Test-Time Domain Adaptation by Learning Domain-Aware Batch Normalization

Trust-aware Conditional Adversarial Domain Adaptation with Feature Norm Alignment.

Class-Level Adaptation Network with Self Training for Unsupervised Domain Adaptation

Revisiting Batch Normalization for Practical Domain Adaptation.

Adaptive Batch Normalization for Practical Domain Adaptation

Meta-TTT: A Meta-learning Minimax Framework For Test-Time Training

DomainAdaptor: A Novel Approach to Test-time Adaptation

Test-time Batch Statistics Calibration for Covariate Shift

Prepare for the Worst: Generalizing across Domain Shifts with Adversarial Batch Normalization

Learning Instance-Specific Adaptation for Cross-Domain Segmentation

InterBN: Channel Fusion for Adversarial Unsupervised Domain Adaptation.

Multiple Source Domain Adaptation with Adversarial Training of Neural Networks

Channel-Selective Normalization for Label-Shift Robust Test-Time Adaptation

Patch-aware Batch Normalization for Improving Cross-domain Robustness

Overcoming Recency Bias of Normalization Statistics in Continual Learning: Balance and Adaptation

Unraveling Batch Normalization for Realistic Test-Time Adaptation

Deep Asymmetric Transfer Network For Unbalanced Domain Adaptation

Semi-Supervised Domain Adaptation for Multi-Label Classification on Nonintrusive Load Monitoring

A Two-Stage Multi-Target Domain Adaptation Framework for Prediction of Key Performance Indicators Based on Adversarial Network

Adaptive Mutual Learning for Unsupervised Domain Adaptation

Blending-target Domain Adaptation by Adversarial Meta-Adaptation Networks