MVDG: Efficient Multi-view 3D Disambiguation on Unconstrained Real-World Images

2026-10-01

AI Summary

This study addresses the "twin" ambiguity arising from similar 3D surfaces in real images and the high inference complexity of conventional pairwise approaches by proposing a scalable multi-view disambiguation framework based on VGGT. The method introduces 3D-aware features to mitigate pairwise dependencies, enabling joint reasoning over an arbitrary number of views through a single encoder-decoder forward pass. Furthermore, a pseudo-pairwise training set is constructed to facilitate stable fine-tuning, and its predictive correlation with Structure-from-Motion (SfM) metrics is established to guide evaluation. Experimental results demonstrate that the proposed framework preserves pairwise matching accuracy while substantially improving both SfM reconstruction quality and inference efficiency, consistently outperforming existing baseline methods.

Abstract

Illusory matches between distinct yet visually similar 3D surfaces--doppelgangers--remain a fundamental obstacle for large-scale, in-the-wild 3D reconstruction and visual localization. Prior work mitigates this issue with pairwise classifiers, but this design limits multi-view contextual reasoning and incurs O(n^2) inference complexity for downstream structure-from-motion (SfM). We present MVDG, a scalable multi-view disambiguation framework built on the 3D foundation model VGGT, which jointly reasons over an arbitrary number of multiview images. By incorporating 3D-aware multi-view features, our method reduces dependence on pairwise comparisons by encoding and decoding views in a single pass. We further observe that direct multi-view fine-tuning of VGGT can be unstable under noisy supervision; motivated by label ambiguity in Doppelgangers, we construct a pseudo-pairwise training set from AerialMegaDepth and show that fine-tuning on sampled subsets yields stable optimization and strong generalization to held-out scenes. Finally, because full SfM evaluation (even with faster pipelines such as GLOMAP) remains expensive, we process a pseudo-pairwise dataset for efficient validation; we derive a predictive relationship between regular SfM metrics and the classification accuracy on this pseudo-pairwise test. Experiments show that our method achieves comparable pairwise accuracy while improving both SfM accuracy and inference speed over baselines.

Problem

• 3D reconstruction
• visual localization
• doppelgangers
• multi-view disambiguation
• structure-from-motion

Innovation

• Multi-view 3D Disambiguation
• 3D Foundation Model
• Structure-from-Motion
• Pseudo-pairwise Training
• Doppelgangers