Consistent Quantity–Quality Control across Scenes for Deployment-Aware Gaussian Splatting

Fengdi Zhang1,2, Hongkun Cao2, Ruqi Huang1
1Tsinghua University, 2Pengcheng Laboratory
arXiv Code Results

TL;DR: ControlGS introduces a cross-scene quantity-quality control hyperparameter, 𝜆𝛼, enabling automatic generation of optimized 3DGS scenes based on semantic compression preferences, suitable for cross-scene and cross-device automated deployment.

Interactive Demo

Adjust 𝜆𝛼 (Lambda) or use “Auto” to auto-select the optimal model based on device performance.

Note: Browser rendering of 3DGS scenes may differ. Refer to the official viewer for accurate results.

Method Overview

Starting from a sparse point cloud reconstructed via SfM, ControlGS initialize an anisotropic Gaussian set and alternate between uniform octree-style subdivision and sparsity-driven pruning:

  • Uniform Gaussian Branching: Whenever the number of pruned Gaussians falls below a threshold, we split all surviving Gaussians into eight children in an octree fashion, inheriting parent attributes to achieve a coarse-to-fine frequency progression without scene-specific heuristics.
  • Gaussian Atrophy: We add an L1 opacity regularization that applies a constant negative gradient to each Gaussian’s opacity, causing low-contribution Gaussians to “self-shrink” and be removed once their opacity drops below a threshold.

The controllable core of ControlGS is a single hyperparameter, 𝜆𝛼, which scales the atrophy loss:

  • Increasing 𝜆𝛼 strengthens pruning, yielding more compact models;
  • Decreasing 𝜆𝛼 preserves more Gaussians for higher-fidelity rendering.

By selecting a 𝜆𝛼 reflecting the quantity–quality preference and training once under a fixed setup, ControlGS enables consistent, wide-range, stepless trade-off control between Gaussian quantity and rendering quality, facilitating the efficient generation of multiple model variants tailored to diverse deployment needs.

Results

Quantity–Quality Control Performance

ControlGS achieves smooth, stepless, and predictable control over the trade-off between rendering quality and Gaussian quantity across diverse scenes, including high-fidelity reconstructions and highly compressed models, and significantly outperforms baseline methods in control consistency, range, and precision.


Comparison with SOTA method

Compared to existing methods, ControlGS achieves higher rendering quality with fewer Gaussians on unseen test views, consistently preserving intricate structures and high-frequency textures across diverse scenes.


BibTeX

      @article{zhang2025consistent,
        title={Consistent Quantity-Quality Control across Scenes for Deployment-Aware Gaussian Splatting},
        author={Zhang, Fengdi and Cao, Hongkun and Huang, Ruqi},
        journal={arXiv preprint arXiv:2505.10473},
        year={2025}
      }