Reproduce “PRIMU: Uncertainty Estimation for Novel Views in Gaussian Splatting from Primitive-Based Representations of Error and Coverage” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Thomas Gottwald, Edgar Heinert, Peter Stehr, Chamuditha Jayanga Galappaththige, Matthias Rottmann
PRIMU is a post-hoc framework for Gaussian Splatting that quantifies uncertainty for novel views by building primitive-based representations of error and coverage directly from training data. It demonstrates improved performance over baselines in both rendering and depth uncertainty estimation, and effectively enhances active view selection for 3D scene reconstruction.
Key insights
- Leveraging primitive-level representations derived from training-view coverage and reconstruction error provides more interpretable and accurate uncertainty information than relying solely on Gaussian parameter variances.
- A post-hoc meta-regression approach enables effective uncertainty prediction without altering the core GS rendering or optimization pipeline, enhancing flexibility and efficiency.
- Uncertainty estimation performance is significantly better for object-centric regions, and meta-regression models can generalize effectively to unseen scenes, expanding applicability.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.