Category-Adaptive Domain Adaptation for Semantic Segmentation

TL;DR

This paper introduces a category-adaptive domain adaptation method for semantic segmentation that uses adversarial style alignment and category-wise pseudo label selection to improve cross-domain performance without target labels.

Contribution

It proposes a novel category-adaptive threshold mechanism and adversarial style bridging to enhance unsupervised domain adaptation for semantic segmentation.

Findings

Outperforms state-of-the-art methods on GTA5 to Cityscapes adaptation

Achieves significant improvement in cross-domain segmentation accuracy

Effectively balances pseudo labels across categories

Abstract

Unsupervised domain adaptation (UDA) becomes more and more popular in tackling real-world problems without ground truth of the target domain. Though tedious annotation work is not required, UDA unavoidably faces two problems: 1) how to narrow the domain discrepancy to boost the transferring performance; 2) how to improve pseudo annotation producing mechanism for self-supervised learning (SSL). In this paper, we focus on UDA for semantic segmentation task. Firstly, we introduce adversarial learning into style gap bridging mechanism to keep the style information from two domains in the similar space. Secondly, to keep the balance of pseudo labels on each category, we propose a category-adaptive threshold mechanism to choose category-wise pseudo labels for SSL. The experiments are conducted using GTA5 as the source domain, Cityscapes as the target domain. The results show that our model outperforms the state-of-the-arts with a noticeable gain on cross-domain adaptation tasks.