Towards Unsupervised Domain Bridging via Image Degradation in Semantic Segmentation

TL;DR

DiDA introduces a novel unsupervised domain bridging method for semantic segmentation that uses image degradation to create intermediate domains and a diffusion encoder to preserve semantic features, leading to improved cross-domain performance.

Contribution

The paper proposes DiDA, a new approach combining degradation-based intermediate domain construction and semantic shift compensation, enhancing unsupervised domain adaptation in semantic segmentation.

Findings

Consistently improves performance across multiple benchmarks.

Supports various degradation operations as a plug-and-play module.

Effectively preserves semantic features during domain adaptation.

Abstract

Semantic segmentation suffers from significant performance degradation when the trained network is applied to a different domain. To address this issue, unsupervised domain adaptation (UDA) has been extensively studied. Despite the effectiveness of selftraining techniques in UDA, they still overlook the explicit modeling of domain-shared feature extraction. In this paper, we propose DiDA, an unsupervised domain bridging approach for semantic segmentation. DiDA consists of two key modules: (1) Degradation-based Intermediate Domain Construction, which creates continuous intermediate domains through simple image degradation operations to encourage learning domain-invariant features as domain differences gradually diminish; (2) Semantic Shift Compensation, which leverages a diffusion encoder to disentangle and compensate for semantic shift information with degraded timesteps, preserving discriminative representations in the intermediate domains. As a plug-and-play solution, DiDA supports various degradation operations and seamlessly integrates with existing UDA methods. Extensive experiments on multiple domain adaptive semantic segmentation benchmarks demonstrate that DiDA consistently achieves significant performance improvements across all settings. Code is available at https://github.com/Woof6/DiDA.