Contrast, Stylize and Adapt: Unsupervised Contrastive Learning Framework for Domain Adaptive Semantic Segmentation

Tianyu Li, Subhankar Roy, Huayi Zhou, Hongtao Lu, Stephane Lathuiliere
6/15/2023
cs.CV

Official code: cxa9264/CONFETI (10 stars, last commit 2023-12-11). 1 more implementation below.

Abstract

To overcome the domain gap between synthetic and real-world datasets, unsupervised domain adaptation methods have been proposed for semantic segmentation. Majority of the previous approaches have attempted to reduce the gap either at the pixel or feature level, disregarding the fact that the two components interact positively. To address this, we present CONtrastive FEaTure and pIxel alignment (CONFETI) for bridging the domain gap at both the pixel and feature levels using a unique contrastive formulation. We introduce well-estimated prototypes by including category-wise cross-domain information to link the two alignments: the pixel-level alignment is achieved using the jointly trained style transfer module with the prototypical semantic consistency, while the feature-level alignment is enforced to cross-domain features with the \textbf{pixel-to-prototype contrast}. Our extensive experiments demonstrate that our method outperforms existing state-of-the-art methods using DeepLabV2. Our code is available at https://github.com/cxa9264/CONFETI

View on arXivView PDF

Code Implementations(2)

cxa9264/CONFETIOfficial100%

Official implementation of Contrast, Stylize and Adapt: Unsupervised Contrastive Learning Framework for Domain Adaptive Semantic Segmentation

101Shell, PythonApr 7, 20232 years ago
cxa9264/CONFETIIOfficial100%

Contrast, Stylize and Adapt: Unsupervised Contrastive Learning Framework for Domain Adaptive Semantic Segmentation

50Shell, PythonDec 11, 20232 years agoNOASSERTION

Similar Papers

Weakly-Supervised Domain Adaptive Semantic Segmentation with Prototypical Contrastive Learning

91% match2023

I2F: A Unified Image-to-Feature Approach for Domain Adaptive Semantic Segmentation

90% match2023

Bi-directional Contrastive Learning for Domain Adaptive Semantic Segmentation

90% match1 repo442022

Prototypical Contrast Adaptation for Domain Adaptive Semantic Segmentation

89% match1 repo612022

FDA: Fourier Domain Adaptation for Semantic Segmentation

87% match1 repo5602020

DynAlign: Unsupervised Dynamic Taxonomy Alignment for Cross-Domain Segmentation

87% match1 repo72025

Cite this paper

@article{li2023contrast,
  title  = {Contrast, Stylize and Adapt: Unsupervised Contrastive Learning Framework for Domain Adaptive Semantic Segmentation},
  author = {Tianyu Li and Subhankar Roy and Huayi Zhou and Hongtao Lu and Stephane Lathuiliere},
  year   = {2023},
  eprint = {2306.09098},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2306.09098v1}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2306.09098.svg)](https://codeofpaper.com/paper/2306.09098)

Discussion