Skip to content
Conference Open access

MarCon: Max-Margin Contrastive Learning for Imbalanced Domain Adaptation Semantic Segmentation

Sep 2026 · Proceedings of the Thirty-Fifth International Joint Conference on Artificial Intelligence · 0 citations · 45 references

TL;DR

This work derives a reformulated pixel-level contrastive learning objective by modeling feature distributions with the von Mises-Fisher distribution, and introduces a Reliability-aware Filter based on the vMF-derived metrics, which performs adaptive class-wise pixel reliability assessment to identify unreliable pixels and attenuate their contribution to model training, thereby mitigating confirmation bias.

Abstract

Unsupervised Domain Adaptation for Semantic Segmentation (UDA-SS) has seen significant progress in recent years. Existing UDA-SS approaches mostly adopt a pseudo-labeling schema to adapt model in the target domain, but they often overlook the inherent long-tailed data distribution in segmentation. We find that such scarce tail samples can lead to representation collapse for tail classes and further hinder the quality of pseudo-labels. To address these challenges, we propose MarCon, a framework that mitigates the long-tailed problem from both feature representation and pseudo-labeling perspectives. Specifically, to explicitly learn a Maximum-Margin Distribution, we derive a reformulated pixel-level contrastive learning objective by modeling feature distributions with the von Mises-Fisher (vMF) distribution. It enforces strict margins to enhance intra-class compactness and inter-class separability, preventing tail classes from being overwhelmed by head classes. Furthermore, to mitigate label noise, we introduce a Reliability-aware Filter (RaF) based on the vMF-derived metrics, which performs adaptive class-wise pixel reliability assessment to identify unreliable pixels and attenuate their contribution to model training, thereby mitigating confirmation bias. Extensive experiments on GTA → Cityscapes and SYNTHIA → Cityscapes demonstrate that MarCon consistently outperforms current leading methods across various transformer-based architectures.

Read PDF

Similar papers

Open access Aug 2026

Contextrast++: Robust Multi-Scale Contextual Contrastive Learning for Semantic Segmentation.

Semantic segmentation has rapidly advanced with deep learning; however, challenges remain in effectively capturing local and global contexts as well as addressing the long-tailed distribution problem. To tackle these issues, we present Contextrast++, a robust contrastive learning method for semantic segmentation that i...

Chan-Yong Sung, Hyungtae Lim, Wanhee Kim et al. · 0 citations
Aug 2026

SELECT: SELEctive Context Transfer for Class-Incremental Semantic Segmentation

This work proposes SELECT, a novel approach for Selective Context Transfer, which instead grounds each new class in a small set of semantically similar past classes, and adds a controlled noise perturbation and a margin-transfer loss that enforces separation between the new class token and its source tokens.

Avi Gupta, Saurabh Yadav, Koteswar Rao Jerripothula et al. · 0 citations
Preprint Aug 2026

Difficulty-Aware Sample Allocation for Adaptive Data Augmentation in Semantic Segmentation

Data augmentation is a standard component of modern semantic segmentation pipelines, but most augmentation techniques allocate transformations uniformly across training samples or adapt to a single difficulty signal such as loss. This ignores the fact that segmentation difficulty is multi-factorial, since ambiguous pre...

O. Arigbabu, Abimbola Ismail Arigbabu · 0 citations
Sep 2026

DR.WILSS: Diffusion-Based Replay for Weakly Supervised Continual Semantic Segmentation

This work introduces DR.WILSS, an innovative approach to address catastrophic forgetting in continual learning using diffusion-based generative replay, which leverages language clues to guide the diffusion process, employing self-inpainting and regularization techniques to efficiently produce replay data, aiding the le...

L. Marx, Francesco Barbato, Matteo Caligiuri et al. · 0 citations
Preprint Aug 2026

LASA: Language-and-Source-Anchored Alignment for Domain Generalized Semantic Segmentation

The Language-and-Source-Anchored Alignment (LASA) framework is proposed, which comprises three synergistic components: Text-and-Source-Guided Style Transfer (TSGST), Domain-Aware Query Adapter (DAQA), and Domain-Aware Decoder Optimizer (DADO).

Jin-Hong Zhu, Wei-Qi Yan, Sheng-Chuan Zhang et al. · 0 citations
Sep 2026

Preference-Guided Adaptation for Open-Vocabulary Semantic Segmentation via Prompt Disagreement

Open-vocabulary semantic segmentation (OVSS) enables pixel-level prediction over arbitrary text-specified vocabularies and has shown strong generalization on common benchmarks. However, OVSS performance often degrades in specialized domains such as medical imaging, remote sensing, and industrial inspection, where dense...

Hyun-Kurl Jang, Jihun Kim, Kuk-Jin Yoon · 0 citations

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.