Speedy-Splat: Fast 3D Gaussian Splatting with Sparse Pixels and Sparse Primitives

TL;DR

Speedy-Splat significantly accelerates 3D Gaussian Splatting rendering by optimizing Gaussian localization and pruning, reducing model size and training time, enabling real-time scene rendering in resource-limited environments.

Contribution

The paper introduces a novel approach combining precise Gaussian localization and pruning techniques to enhance 3D-GS rendering speed and efficiency.

Findings

6.71x average rendering speed increase across multiple datasets

Reduced model size and training time

Maintained visual fidelity with faster rendering

Abstract

3D Gaussian Splatting (3D-GS) is a recent 3D scene reconstruction technique that enables real-time rendering of novel views by modeling scenes as parametric point clouds of differentiable 3D Gaussians. However, its rendering speed and model size still present bottlenecks, especially in resource-constrained settings. In this paper, we identify and address two key inefficiencies in 3D-GS to substantially improve rendering speed. These improvements also yield the ancillary benefits of reduced model size and training time. First, we optimize the rendering pipeline to precisely localize Gaussians in the scene, boosting rendering speed without altering visual fidelity. Second, we introduce a novel pruning technique and integrate it into the training pipeline, significantly reducing model size and training time while further raising rendering speed. Our Speedy-Splat approach combines these techniques to accelerate average rendering speed by a drastic $\mathit{6.71\times}$ across scenes from the Mip-NeRF 360, Tanks & Temples, and Deep Blending datasets.