---
title: "MarCon: Max-Margin Contrastive Learning for Imbalanced Domain Adaptation Semantic Segmentation"
authors:
  - "Yibo Wang"
  - "Ruikang Xu"
  - "Guangcheng Zhu"
  - "Cheng Peng"
  - "Haobo Wang"
  - "Runze Wu"
  - "Minmin Lin"
  - "Changjie Fan"
doi: "10.24963/ijcai.2026/556"
url: "https://ijcai.org/proceedings/2026/556"
pdf: "https://ijcai.org/proceedings/2026/0556.pdf"
venue: "Proceedings of the Thirty-Fifth International Joint Conference on Artificial Intelligence"
year: 2026
pages: "4994-5002"
track: "Main Track"
keywords:
  - "Machine Learning: Weakly supervised learning"
---

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

Yibo Wang, Ruikang Xu, Guangcheng Zhu, Cheng Peng, Haobo Wang, Runze Wu, Minmin Lin, Changjie Fan

*Proceedings of the Thirty-Fifth International Joint Conference on Artificial Intelligence*, pages 4994-5002. DOI: [10.24963/ijcai.2026/556](https://doi.org/10.24963/ijcai.2026/556)

## 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.