LeGS turns density control in 3D Gaussian Splatting into a learnable RL policy whose reward is derived from a closed-form sensitivity analysis that measures each Gaussian's marginal contribution to reconstruction quality.
Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
fields
cs.CV 2years
2026 2verdicts
UNVERDICTED 2representative citing papers
SIG frequency scheduler and sphere-constrained Gaussians enable more efficient and higher-quality 3D Gaussian Splatting for large-scale scenes by synchronizing supervision with representation frequencies.
citing papers explorer
-
Beyond Heuristics: Learnable Density Control for 3D Gaussian Splatting
LeGS turns density control in 3D Gaussian Splatting into a learnable RL policy whose reward is derived from a closed-form sensitivity analysis that measures each Gaussian's marginal contribution to reconstruction quality.
-
Structure-Aware Gaussian Splatting for Large-Scale Scene Reconstruction
SIG frequency scheduler and sphere-constrained Gaussians enable more efficient and higher-quality 3D Gaussian Splatting for large-scale scenes by synchronizing supervision with representation frequencies.