Tile skipper speeds up 3d gaussian splatting with adaptive pruning
TileSkipper: Region-Adaptive Tile Pruning for 3D Gaussian Splatting
Computer Vision and Pattern Recognition
Summary
Rendering 3D scenes using Gaussian splats often wastes time by treating all parts of the scene the same way. The authors developed TileSkipper, which picks different shortcuts for parts of the scene based on their visual importance. This approach speeds up rendering by skipping unnecessary calculations while keeping the image quality nearly the same. It works by pre-calculating where to cut off support for each Gaussian and using a small extra data amount for efficient rendering.
What this means in practice
- •For graphics programmers: Optimize 3D Gaussian splatting renderers by using per-Gaussian cutoff policies for faster tile pruning without additional runtime overhead.
- •For game engine developers: Improve real-time rendering speeds in games using Gaussian splatting by integrating region-adaptive tile pruning techniques.
Authors
Jingxing Li, Yongjae Lee, Deliang Fan, Abhay Kumar Yadav, Cheng Peng, Rama Chellappa
Abstract
Tiled 3D Gaussian Splatting rasterizers often use one scene-wide contribution cutoff for tile enumeration, although content differs in its sensitivity to support truncation. TileSkipper selects a static per-Gaussian cutoff policy for a frozen checkpoint. Calibration renders measure candidate pair savings and an isolated-removal distortion proxy that accounts for front transmittance and background color. The method allocates cutoffs across 64 Gaussian groups and accepts policies only after complete renders on disjoint selection views. The exported policy uses one byte per Gaussian, with no parameter updates, additional kernel, or per-frame policy inference. Across 13 scenes from Mip-NeRF 360, Tanks & Temples, and Deep Blending, a fixed-policy AccuTile sweep gives dataset-macro speedups of $1.088\times$ at standard resolution and $1.238\times$ at 3840 pixels wide, with $-0.007/-0.023$ dB mean PSNR change. Six integrations with existing opacity-aware bounds yield $1.009\times$--$1.121\times$ compiler-only speedups. For four ports from $3σ$ rasterizers, we separately attribute the prior exact-bound transition and our incremental gain. Matched-quality ablations show modest gains over scene-global calibration and parity with per-Gaussian control; the standalone comparison with AdaGScale is regime-dependent.