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

TL;DR

This paper introduces a test-time domain adaptation method that manipulates only the affine parameters of batch normalization layers, combined with self-supervised learning and meta-learning, to improve adaptation to unseen domains without increasing inference cost.

Contribution

The work proposes a novel approach focusing on updating only BN affine parameters and using an auxiliary SSL branch with meta-learning for effective domain adaptation.

Findings

Outperforms prior methods on five WILDS datasets.

Maintains same inference cost by discarding auxiliary branch after adaptation.

Enhances domain knowledge extraction through self-supervised learning.

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 https://github.com/ynanwu/MABN.