Learning Fast 3D Gaussian Splatting Rendering using Continuous Level of Detail

Abstract

3D Gaussian splatting (3DGS) has shown potential for rendering photorealistic 3D scenes in real-time, but rendering these scenes on less powerful hardware is still a challenge, especially with high-resolution displays. We introduce a continuous level of detail (CLOD) algorithm that learns to order splats based on importance and optimizes them so that a representative and realistic scene can be rendered for an arbitrary splat count. Our method requires no additional memory or rendering overhead, works with existing 3DGS renderers, and extends to distance-based LOD selection, foveated rendering, and budget-based rendering.