Matryoshka Gaussian Splatting

TL;DR

Matryoshka Gaussian Splatting (MGS) introduces a training framework that enables continuous level of detail in 3D Gaussian Splatting, allowing smooth quality scaling without sacrificing full-capacity rendering performance.

Contribution

MGS is a novel training method that learns an ordered set of Gaussians, enabling continuous LoD in 3D Gaussian Splatting without architectural changes.

Findings

MGS matches full-capacity performance across benchmarks.

Enables smooth speed-quality trade-off from a single model.

Requires only two forward passes per iteration.

Abstract

The ability to render scenes at adjustable fidelity from a single model, known as level of detail (LoD), is crucial for practical deployment of 3D Gaussian Splatting (3DGS). Existing discrete LoD methods expose only a limited set of operating points, while concurrent continuous LoD approaches enable smoother scaling but often suffer noticeable quality degradation at full capacity, making LoD a costly design decision. We introduce Matryoshka Gaussian Splatting (MGS), a training framework that enables continuous LoD for standard 3DGS pipelines without sacrificing full-capacity rendering quality. MGS learns a single ordered set of Gaussians such that rendering any prefix, the first k splats, produces a coherent reconstruction whose fidelity improves smoothly with increasing budget. Our key idea is stochastic budget training: each iteration samples a random splat budget and optimises both the corresponding prefix and the full set. This strategy requires only two forward passes and introduces no architectural modifications. Experiments across four benchmarks and six baselines show that MGS matches the full-capacity performance of its backbone while enabling a continuous speed-quality trade-off from a single model. Extensive ablations on ordering strategies, training objectives, and model capacity further validate the designs.