SpotlessSplats: Ignoring Distractors in 3D Gaussian Splatting

TL;DR

SpotLessSplats enhances 3D Gaussian Splatting by effectively ignoring transient distractors using pre-trained features and robust optimization, enabling high-quality real-world 3D reconstructions in casual settings.

Contribution

The paper introduces SpotLessSplats, a novel method that improves 3D Gaussian Splatting by handling distractors, allowing for better real-world 3D reconstruction without controlled environments.

Findings

Achieves state-of-the-art reconstruction quality on casual captures.

Effectively ignores transient distractors in real-world scenes.

Compatible with existing 3D Gaussian Splatting frameworks.

Abstract

3D Gaussian Splatting (3DGS) is a promising technique for 3D reconstruction, offering efficient training and rendering speeds, making it suitable for real-time applications.However, current methods require highly controlled environments (no moving people or wind-blown elements, and consistent lighting) to meet the inter-view consistency assumption of 3DGS. This makes reconstruction of real-world captures problematic. We present SpotLessSplats, an approach that leverages pre-trained and general-purpose features coupled with robust optimization to effectively ignore transient distractors. Our method achieves state-of-the-art reconstruction quality both visually and quantitatively, on casual captures. Additional results available at: https://spotlesssplats.github.io