Pixel-Level Cycle Association: A New Perspective for Domain Adaptive Semantic Segmentation

TL;DR

This paper introduces a pixel-level cycle association approach for domain adaptive semantic segmentation, emphasizing pixel-wise relationships to better align source and target domains, outperforming previous global discrepancy methods.

Contribution

It proposes a novel pixel-level cycle association framework that enhances feature discrimination and domain alignment without extra parameters, advancing domain adaptive segmentation.

Findings

Outperforms previous state-of-the-art methods on GTAV and SYNTHIA benchmarks.

Effective end-to-end training without additional parameters.

Demonstrates the importance of pixel-wise relationships in domain adaptation.

Abstract

Domain adaptive semantic segmentation aims to train a model performing satisfactory pixel-level predictions on the target with only out-of-domain (source) annotations. The conventional solution to this task is to minimize the discrepancy between source and target to enable effective knowledge transfer. Previous domain discrepancy minimization methods are mainly based on the adversarial training. They tend to consider the domain discrepancy globally, which ignore the pixel-wise relationships and are less discriminative. In this paper, we propose to build the pixel-level cycle association between source and target pixel pairs and contrastively strengthen their connections to diminish the domain gap and make the features more discriminative. To the best of our knowledge, this is a new perspective for tackling such a challenging task. Experiment results on two representative domain adaptation benchmarks, i.e. GTAV $\rightarrow$ Cityscapes and SYNTHIA $\rightarrow$ Cityscapes, verify the effectiveness of our proposed method and demonstrate that our method performs favorably against previous state-of-the-arts. Our method can be trained end-to-end in one stage and introduces no additional parameters, which is expected to serve as a general framework and help ease future research in domain adaptive semantic segmentation. Code is available at https://github.com/kgl-prml/Pixel- Level-Cycle-Association.