2025 Project

SFMambaNet: Spectral-Frequency Correspondence Pruning

A spectral-frequency enhanced selective state space model for efficient visual correspondence pruning.

Computer VisionState Space ModelsMambaCorrespondence PruningFrequency DomainGraph Neural NetworksTwo-View GeometryCamera Pose EstimationOutlier Rejection
arXiv / DOI https://doi.org/10.48550/arXiv.2606.04493

SFMambaNet investigates efficient global context modeling for correspondence pruning.

Paper labels: SCI Zone 1; CCF B.

SFMambaNet framework

The project introduces spectral-frequency awareness into a selective state space architecture. This direction aims to avoid the quadratic scaling of Transformer-style global attention while keeping strong global modeling capacity.

My work included literature review, idea formulation, experiments, and manuscript writing under faculty supervision.