English

Modeling Uncertainty in 3D Gaussian Splatting through Continuous Semantic Splatting

Robotics 2024-11-06 v1 Computer Vision and Pattern Recognition

Abstract

In this paper, we present a novel algorithm for probabilistically updating and rasterizing semantic maps within 3D Gaussian Splatting (3D-GS). Although previous methods have introduced algorithms which learn to rasterize features in 3D-GS for enhanced scene understanding, 3D-GS can fail without warning which presents a challenge for safety-critical robotic applications. To address this gap, we propose a method which advances the literature of continuous semantic mapping from voxels to ellipsoids, combining the precise structure of 3D-GS with the ability to quantify uncertainty of probabilistic robotic maps. Given a set of images, our algorithm performs a probabilistic semantic update directly on the 3D ellipsoids to obtain an expectation and variance through the use of conjugate priors. We also propose a probabilistic rasterization which returns per-pixel segmentation predictions with quantifiable uncertainty. We compare our method with similar probabilistic voxel-based methods to verify our extension to 3D ellipsoids, and perform ablation studies on uncertainty quantification and temporal smoothing.

Keywords

Cite

@article{arxiv.2411.02547,
  title  = {Modeling Uncertainty in 3D Gaussian Splatting through Continuous Semantic Splatting},
  author = {Joey Wilson and Marcelino Almeida and Min Sun and Sachit Mahajan and Maani Ghaffari and Parker Ewen and Omid Ghasemalizadeh and Cheng-Hao Kuo and Arnie Sen},
  journal= {arXiv preprint arXiv:2411.02547},
  year   = {2024}
}
R2 v1 2026-06-28T19:48:04.808Z