Seeing Beyond: Extrapolative Domain Adaptive Panoramic Segmentation

2026-03-16

AI Summary

This work addresses the performance degradation in cross-domain panoramic semantic segmentation caused by geometric distortions from field-of-view variations and semantic inconsistencies under open-set conditions. To tackle this, the authors propose the EDA-PSeg framework, which establishes the first open-set domain adaptive panoramic segmentation setting. The method trains on local perspective views and evaluates on 360° panoramic images, leveraging Eulerian Boundary Attention (EMA) to learn viewpoint-invariant representations. It further integrates a higher-order Graph Matching Adapter (GMA) to align shared semantics while decoupling the structural patterns of unknown categories. Evaluated on four benchmark datasets, EDA-PSeg achieves state-of-the-art performance, significantly enhancing model robustness and generalization across viewpoint changes and environmental shifts.

Technology Category

• Computer Vision: Segmentation
• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

Cross-domain panoramic semantic segmentation has attracted growing interest as it enables comprehensive 360° scene understanding for real-world applications. However, it remains particularly challenging due to severe geometric Field of View (FoV) distortions and inconsistent open-set semantics across domains. In this work, we formulate an open-set domain adaptation setting, and propose Extrapolative Domain Adaptive Panoramic Segmentation (EDA-PSeg) framework that trains on local perspective views and tests on full 360° panoramic images, explicitly tackling both geometric FoV shifts across domains and semantic uncertainty arising from previously unseen classes. To this end, we propose the Euler-Margin Attention (EMA), which introduces an angular margin to enhance viewpoint-invariant semantic representation, while performing amplitude and phase modulation to improve generalization toward unseen classes. Additionally, we design the Graph Matching Adapter (GMA), which builds high-order graph relations to align shared semantics across FoV shifts while effectively separating novel categories through structural adaptation. Extensive experiments on four benchmark datasets under camera-shift, weather-condition, and open-set scenarios demonstrate that EDA-PSeg achieves state-of-the-art performance, robust generalization to diverse viewing geometries, and resilience under varying environmental conditions. The code is available at https://github.com/zyfone/EDA-PSeg.

Problem

• domain adaptation
• panoramic segmentation
• field of view distortion
• open-set semantics
• cross-domain

Innovation

• Extrapolative Domain Adaptation
• Panoramic Semantic Segmentation
• Euler-Margin Attention
• Graph Matching Adapter
• Open-Set Recognition