Stable mask pruning reduces 3D Gaussian data for better view synthesis
LinearMask-GS: Stable-Mask Importance Pruning for Compact 3D Gaussian Splatting
Computer Vision and Pattern Recognition
Summary
Creating 3D scenes from many small points can take up a lot of space. The authors found that current methods for trimming unnecessary points often make quick, unreliable decisions. They introduced a new way to gradually decide which points to keep, leading to fewer points without losing image quality. This makes 3D scenes smaller and faster to work with.
What this means in practice
- •For graphics engineers: Reduce the number of 3D Gaussian primitives in rendering pipelines while maintaining or improving image quality for real-time novel view synthesis.
- •For game developers: Create more compact 3D scene representations that lower storage needs and enhance rendering performance in outdoor virtual environments.
Authors
Donghun Ryu, Minhyeok Lee
Abstract
3D Gaussian Splatting (3DGS) enables real-time novel view synthesis but produces millions of primitives through adaptive densification, leading to significant storage overhead. Learned-mask pruning methods such as LP-3DGS address this by assigning each Gaussian a learnable mask to identify and prune redundant primitives. However, we identify a limitation of this paradigm: the steep slope of the Gumbel-Sigmoid activation drives mask values to the extremes within the short mask-training window, before the importance ranking has stabilized, producing a sharply bimodal distribution from which that ranking can no longer be reliably recovered. We propose LinearMask-GS, which replaces Gumbel-Sigmoid with a linear increment activation that keeps mask values in a mid-confidence regime throughout mask training, producing a stable, unimodal mask distribution whose ranking tracks importance. On Mip-NeRF 360, our method achieves 3.6x and 1.6x Gaussian reductions over 3DGS and LP-3DGS, respectively, while maintaining or improving rendering quality. For outdoor scenes, it yields a 1.6x reduction (from 2.18M to 1.36M) with notable gains in PSNR (+0.38 dB), SSIM (+0.025), and LPIPS (-0.029).