SegDA: Maximum Separable Segment Mask with Pseudo Labels for Domain Adaptive Semantic Segmentation

TL;DR

SegDA introduces a novel domain adaptation approach for semantic segmentation that leverages maximum separable segment representations and pseudo label noise correction, significantly improving performance across various benchmarks.

Contribution

The paper proposes a new architecture using ETF classifiers and noise estimation to enhance domain adaptation in semantic segmentation, addressing class confusion and label noise.

Findings

Outperforms existing methods with +2.2 mIoU on GTA to Cityscapes

Achieves +2.0 mIoU on Synthia to Cityscapes

Improves +5.9 mIoU on Cityscapes to DarkZurich

Abstract

Unsupervised Domain Adaptation (UDA) aims to solve the problem of label scarcity of the target domain by transferring the knowledge from the label rich source domain. Usually, the source domain consists of synthetic images for which the annotation is easily obtained using the well known computer graphics techniques. However, obtaining annotation for real world images (target domain) require lot of manual annotation effort and is very time consuming because it requires per pixel annotation. To address this problem we propose SegDA module to enhance transfer performance of UDA methods by learning the maximum separable segment representation. This resolves the problem of identifying visually similar classes like pedestrian/rider, sidewalk/road etc. We leveraged Equiangular Tight Frame (ETF) classifier inspired from Neural Collapse for maximal separation between segment classes. This causes the source domain pixel representation to collapse to a single vector forming a simplex vertices which are aligned to the maximal separable ETF classifier. We use this phenomenon to propose the novel architecture for domain adaptation of segment representation for target domain. Additionally, we proposed to estimate the noise in labelling the target domain images and update the decoder for noise correction which encourages the discovery of pixels for classes not identified in pseudo labels. We have used four UDA benchmarks simulating synthetic-to-real, daytime-to-nighttime, clear-to-adverse weather scenarios. Our proposed approach outperforms +2.2 mIoU on GTA -> Cityscapes, +2.0 mIoU on Synthia -> Cityscapes, +5.9 mIoU on Cityscapes -> DarkZurich, +2.6 mIoU on Cityscapes -> ACDC.