Consistency Regularization with High-dimensional Non-adversarial Source-guided Perturbation for Unsupervised Domain Adaptation in Segmentation

TL;DR

This paper introduces BiSIDA, a simple yet effective bidirectional style transfer method for unsupervised domain adaptation in segmentation, leveraging consistency regularization and style-induced perturbations to improve performance without complex adversarial training.

Contribution

Proposes a novel bidirectional style transfer approach, BiSIDA, that enhances unsupervised domain adaptation in segmentation by exploiting unlabeled target data with simple neural style transfer.

Findings

Achieves state-of-the-art results on GTA5-to-CityScapes benchmark.

Outperforms existing methods on SYNTHIA-to-CityScapes benchmark.

Utilizes only simple style transfer without adversarial training.

Abstract

Unsupervised domain adaptation for semantic segmentation has been intensively studied due to the low cost of the pixel-level annotation for synthetic data. The most common approaches try to generate images or features mimicking the distribution in the target domain while preserving the semantic contents in the source domain so that a model can be trained with annotations from the latter. However, such methods highly rely on an image translator or feature extractor trained in an elaborated mechanism including adversarial training, which brings in extra complexity and instability in the adaptation process. Furthermore, these methods mainly focus on taking advantage of the labeled source dataset, leaving the unlabeled target dataset not fully utilized. In this paper, we propose a bidirectional style-induced domain adaptation method, called BiSIDA, that employs consistency regularization to efficiently exploit information from the unlabeled target domain dataset, requiring only a simple neural style transfer model. BiSIDA aligns domains by not only transferring source images into the style of target images but also transferring target images into the style of source images to perform high-dimensional perturbation on the unlabeled target images, which is crucial to the success in applying consistency regularization in segmentation tasks. Extensive experiments show that our BiSIDA achieves new state-of-the-art on two commonly-used synthetic-to-real domain adaptation benchmarks: GTA5-to-CityScapes and SYNTHIA-to-CityScapes.