Search papers, labs, and topics across Lattice.
The paper introduces DualSplat, a novel approach to improve 3D Gaussian Splatting (3DGS) in the presence of transient objects that violate multi-view consistency. DualSplat leverages initial reconstruction failures, specifically incomplete fragments of transient objects, to generate pseudo-masks that guide a second, cleaner 3DGS optimization. An MLP refines these masks online, transitioning from prior supervision to self-consistency, leading to improved performance in transient-heavy scenes.
Turn your 3D Gaussian Splatting failures into features: DualSplat uses initial reconstruction artifacts to bootstrap robust scene representations in the presence of transient objects.
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.