Exploring Consistency in Cross-Domain Transformer for Domain Adaptive Semantic Segmentation

TL;DR

This paper introduces a novel attention consistency approach for domain adaptive transformers in semantic segmentation, aligning attention maps across domains and views to improve target domain performance.

Contribution

It proposes a cross-domain attention map consistency method that enhances domain adaptation in transformer-based semantic segmentation models.

Findings

Outperforms state-of-the-art on GTAV-to-Cityscapes, Synthia-to-Cityscapes, and Cityscapes-to-ACDC benchmarks.

Achieves 1.3, 0.6, and 1.1 percentage point improvements respectively.

Demonstrates effectiveness and generalizability through extensive experiments.

Abstract

While transformers have greatly boosted performance in semantic segmentation, domain adaptive transformers are not yet well explored. We identify that the domain gap can cause discrepancies in self-attention. Due to this gap, the transformer attends to spurious regions or pixels, which deteriorates accuracy on the target domain. We propose to perform adaptation on attention maps with cross-domain attention layers that share features between the source and the target domains. Specifically, we impose consistency between predictions from cross-domain attention and self-attention modules to encourage similar distribution in the attention and output of the model across domains, i.e., attention-level and output-level alignment. We also enforce consistency in attention maps between different augmented views to further strengthen the attention-based alignment. Combining these two components, our method mitigates the discrepancy in attention maps across domains and further boosts the performance of the transformer under unsupervised domain adaptation settings. Our model outperforms the existing state-of-the-art baseline model on three widely used benchmarks, including GTAV-to-Cityscapes by 1.3 percent point (pp), Synthia-to-Cityscapes by 0.6 pp, and Cityscapes-to-ACDC by 1.1 pp, on average. Additionally, we verify the effectiveness and generalizability of our method through extensive experiments. Our code will be publicly available.