---
title: "G-SalAlignMamba: Geometry-Aware Vision Mamba for Dual-Modal Salient Object Detection"
authors:
  - "Haixiao Gao"
  - "Yimin Zheng"
  - "Mengke Song"
  - "Linyou Xiao"
  - "Tian-Tian Zhang"
  - "Zhi-Ri Tang"
doi: "10.24963/ijcai.2026/123"
url: "https://ijcai.org/proceedings/2026/123"
pdf: "https://ijcai.org/proceedings/2026/0123.pdf"
venue: "Proceedings of the Thirty-Fifth International Joint Conference on Artificial Intelligence"
year: 2026
pages: "1097-1105"
track: "Main Track"
keywords:
  - "Computer Vision: Multimodal learning"
  - "Computer Vision: Scene analysis and understanding"
---

# G-SalAlignMamba: Geometry-Aware Vision Mamba for Dual-Modal Salient Object Detection

Haixiao Gao, Yimin Zheng, Mengke Song, Linyou Xiao, Tian-Tian Zhang, Zhi-Ri Tang

*Proceedings of the Thirty-Fifth International Joint Conference on Artificial Intelligence*, pages 1097-1105. DOI: [10.24963/ijcai.2026/123](https://doi.org/10.24963/ijcai.2026/123)

## Abstract

Recently, Visual State Space Models offer powerful global modeling for Dual-modal Salient Object Detection (SOD). However, they are still constrained by three inherent limitations: first, Mamba's strict reliance on sequential ordering makes it sensitive to cross-modal geometric misalignment, where spatial shifts disrupt token correspondence; second, general indiscriminate scanning treating all tokens equally may lead to signal dilution, where sparse foreground features are overwhelmed by background noise; third, conventional decoders rely on implicit upsampling, causing boundary degradation during resolution recovery. To address these challenges, we propose G-SalAlignMamba, a geometry-aware framework tailored for dual-modal SOD. We introduce Geometry-Aware Encoding with explicit alignment to correct spatial shifts, Semantics-Informed Refinement to prevent signal dilution by prioritizing foregrounds, and Structure-Preserving Decoding that integrates explicit alignment with unsupervised boundary refinement. Extensive experiments show that G-SalAlignMamba achieves state-of-the-art performance on RGB-D and RGB-T benchmarks with favorable efficiency (30.41 FPS, 83.80M parameters). The code is available at https://github.com/PC1-99/G-SalAlignMamb.git.