MGCA-Net: Multi-Graph Contextual Attention Network for Two-View Correspondence Learning

2025-12-29

AI Summary

Insufficient local geometric modeling and difficulty in cross-stage information optimization in two-view matching lead to inaccurate geometric constraint capture and poor robustness. To address these issues, we propose a dual-module framework: Contextual Geometric Attention (CGA) and Cross-Stage Multi-Graph Consensus (CSMGC). CGA introduces the first dynamic joint modeling of spatial position and feature representations, significantly enhancing local geometric awareness. CSMGC leverages a sparse graph neural network to enforce cross-stage geometric consistency, integrating multi-scale feature interaction with explicit geometric constraints. Our method achieves substantial improvements over state-of-the-art approaches on YFCC100M and SUN3D benchmarks, particularly in outlier rejection and camera pose estimation—demonstrating superior accuracy and robustness under challenging real-world conditions.

Technology Category

• Computer Vision: Multi-modal Vision
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning
• Machine Learning: Multi-instance/Multi-view Learning

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG

Abstract

Two-view correspondence learning is a key task in computer vision, which aims to establish reliable matching relationships for applications such as camera pose estimation and 3D reconstruction. However, existing methods have limitations in local geometric modeling and cross-stage information optimization, which make it difficult to accurately capture the geometric constraints of matched pairs and thus reduce the robustness of the model. To address these challenges, we propose a Multi-Graph Contextual Attention Network (MGCA-Net), which consists of a Contextual Geometric Attention (CGA) module and a Cross-Stage Multi-Graph Consensus (CSMGC) module. Specifically, CGA dynamically integrates spatial position and feature information via an adaptive attention mechanism and enhances the capability to capture both local and global geometric relationships. Meanwhile, CSMGC establishes geometric consensus via a cross-stage sparse graph network, ensuring the consistency of geometric information across different stages. Experimental results on two representative YFCC100M and SUN3D datasets show that MGCA-Net significantly outperforms existing SOTA methods in the outlier rejection and camera pose estimation tasks. Source code is available at http://www.linshuyuan.com.

Problem

• Enhances local and global geometric relationship capture
• Improves cross-stage geometric information consistency
• Boosts outlier rejection and camera pose estimation accuracy

Innovation

• Adaptive attention integrates spatial and feature information
• Cross-stage sparse graph network ensures geometric consistency
• Multi-graph contextual attention enhances local and global relationships