Gaussian splat objects completed accurately using 2D image AI guidance
GSComplete: Gaussian Splat Completion with 2D Diffusion Priors
Computer Vision and Pattern Recognition
Summary
Gaussian splats are a way to store 3D objects quickly and in detail, but often parts are missing when the data is incomplete. The authors designed a method called GSComplete that fills in these missing parts while keeping the original visible pieces unchanged. Instead of needing complex 3D training data, their method uses 2D image AI tools to guide the completion. Tests show GSComplete keeps original parts much better than earlier methods and creates natural-looking filled-in areas.
Gaussian splats3D object representationcompletionScore Distillation Sampling2D diffusion priorspreservation loss3D generationpartial input dataneural networksimage synthesis
Authors
Elias Brugger, Philipp Erler, Stefan Ohrhallinger, Paul Guerrero
Abstract
Gaussian splats provide a fast, high-fidelity representation for 3D objects but are often constructed from incomplete input data in practice, leaving missing regions. Existing completion methods either do not preserve the original splats or require scarcely available 3D training data. We propose GSComplete, which combines 3D generation based on Score Distillation Sampling with a novel preservation loss that encourages the original splats to be preserved where they should be visible. This effectively completes the Gaussian splat object using only 2D diffusion priors while fully preserving existing splats and generating new splats only in missing regions, without occluding the input. To evaluate our approach, we introduce a new dataset of partial Gaussian splat objects and show that GSComplete achieves significantly more accurate preservation of the input than existing methods with comparable plausibility of the completed result. Our code and dataset will be made available upon acceptance.