Smart target point control for Gaussian Splatting methods

TL;DR

This paper introduces a target point control scheme for Gaussian Splatting that ensures fair and capacity-matched evaluation by dynamically adjusting densification and pruning hyper-parameters to reach a quadratic target count trajectory.

Contribution

It proposes a novel quota-governor method that maintains standard densification cadence while achieving a specified point count, enabling fairer comparisons across methods.

Findings

Achieves target point count within 15k iterations without abrupt cutoffs.

Ensures equal densification and pruning across views and methods.

Facilitates fairer, capacity-matched evaluation of Gaussian Splatting methods.

Abstract

Standard Gaussian splatting methods rely on heuristic densification and pruning to adaptively allocate primitives during training, and the resulting Gaussian count strongly influences both reconstruction quality and runtime. This makes comparisons across methods fragile: improvements can stem from higher representational capacity rather than algorithmic design. A common and naive workaround for this is hard-stopping or budgeting densification/pruning once a target count is reached, which biases training because different methods hit the cap at different times, yielding non-uniform densify/prune exposure across views and uneven point distributions. We propose a target point control scheme that preserves the standard densification window and cadence, but adjusts only the existing densification and opacity-culling hyper-parameters to track a quadratic target count trajectory. This quota-governor reaches the desired count by 15k iterations without abrupt cutoffs, ensuring that all methods and views receive equal densification and pruning cycles, enabling fairer, capacity-matched evaluation.