Multi-source Domain Adaptation for Semantic Segmentation

TL;DR

This paper introduces MADAN, a novel multi-source domain adaptation framework for semantic segmentation that aligns multiple synthetic sources with real-world data, improving performance over existing single-source methods.

Contribution

The paper presents a new end-to-end framework, MADAN, for multi-source domain adaptation in semantic segmentation, incorporating dynamic semantic consistency and domain aggregation techniques.

Findings

MADAN outperforms state-of-the-art methods on multiple datasets.

Effective multi-source adaptation improves segmentation accuracy.

Proposed framework is trained end-to-end with competitive results.

Abstract

Simulation-to-real domain adaptation for semantic segmentation has been actively studied for various applications such as autonomous driving. Existing methods mainly focus on a single-source setting, which cannot easily handle a more practical scenario of multiple sources with different distributions. In this paper, we propose to investigate multi-source domain adaptation for semantic segmentation. Specifically, we design a novel framework, termed Multi-source Adversarial Domain Aggregation Network (MADAN), which can be trained in an end-to-end manner. First, we generate an adapted domain for each source with dynamic semantic consistency while aligning at the pixel-level cycle-consistently towards the target. Second, we propose sub-domain aggregation discriminator and cross-domain cycle discriminator to make different adapted domains more closely aggregated. Finally, feature-level alignment is performed between the aggregated domain and target domain while training the segmentation network. Extensive experiments from synthetic GTA and SYNTHIA to real Cityscapes and BDDS datasets demonstrate that the proposed MADAN model outperforms state-of-the-art approaches. Our source code is released at: https://github.com/Luodian/MADAN.