DualSplat: Robust 3D Gaussian Splatting via Pseudo-Mask Bootstrapping from Reconstruction Failures

arXiv:2604.21631v1 Announce Type: new Abstract: While 3D Gaussian Splatting (3DGS) achieves real-time photorealistic rendering, its performance degrades significantly when training images contain transient objects that violate multi-view consistency. Existing methods face a circular dependency: accurate transient detection requires a well-reconstructed static scene, while clean reconstruction itself depends on reliable transient masks. We address this challenge with DualSplat, a Failure-to-Prior framework that converts first-pass reconstruction failures into explicit priors for a second reconstruction stage. We observe that transients, which appear in only a subset of views, often manifest as incomplete fragments during conservative initial training. We exploit these failures to construct object-level pseudo-masks by combining photometric residuals, feature mismatches, and SAM2 instance boundaries. These pseudo-masks then guide a clean second-pass 3DGS optimization, while a lightweight MLP refines them online by gradually shifting from prior supervision to self-consistency. Experiments on RobustNeRF and NeRF On-the-go show that DualSplat outperforms existing baselines, demonstrating particularly clear advantages in transient-heavy scenes and transient regions.

Related

• GSCompleter: A Distillation-Free Plugin for Metric-Aware 3D Gaussian Splatting Completion in Seconds
• PDF-GS: Progressive Distractor Filtering for Robust 3D Gaussian Splatting
• MetroGS: Efficient and Stable Reconstruction of Geometrically Accurate High-Fidelity Large-Scale Scenes
• GaussianFlow SLAM: Monocular Gaussian Splatting SLAM Guided by GaussianFlow
• TransSplat: Unbalanced Semantic Transport for Language-Driven 3DGS Editing

Source: arXiv cs.CV | 2026-04-24