Research map / 3D vision / Gaussian splatting
StochasticSplats: Stochastic Rasterization for Sorting-Free 3D Gaussian Splatting
Shakiba Kheradmand, Delio Vicini, George Kopanas, Dmitry Lagun, Kwang Moo Yi, Mark Matthews, Andrea Tagliasacchi
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| increase samples per pixel from 1 to 1024 (ablation) | StochasticSplats · MipNerf360 (fine-tuned from StopThePop scenes) | PSNR: 17.95 -> 28.66 | Improved |
| increase samples per pixel from 1 to 1024 (ablation) | StochasticSplats · MipNerf360 (fine-tuned from StopThePop scenes) | LPIPS: 0.611 -> 0.168 | Improved |
| increase samples per pixel from 1 to 16 (ablation) | StochasticSplats-OpenGL · MipNerf360 forward pass time (RTX 3090) | forward time (ms): 3.25 -> 18.48 | Hurt |
| use stochastic transparency at 1 SPP instead of sorted alpha blending (OpenGL vs 3DGS-CUDA) | 3DGS · MipNerf360 forward pass time (RTX 3090) | forward time (ms): 8.14 -> 3.25 | Improved |
| port 3DGS from CUDA to OpenGL | 3DGS · MipNerf360 forward pass time (RTX 3090) | forward time (ms): 8.14 -> 32.15 | Hurt |
| use stochastic transparency with 1 SPP in the backward pass | 3DGS · MipNerf360 backward pass time (RTX 3090, CUDA) | backward time (ms): 23.06 -> 18.10 | Improved |
| increase backward samples per pixel from 1 to 128 (ablation) | StochasticSplats · MipNerf360 backward pass time (RTX 3090, CUDA) | backward time (ms): 18.10 -> 478.93 | Hurt |
| replace alpha blending renderer with StochasticSplats at 1 SPP in LangSplat | LangSplat · LERF open-vocab localization (mean) | accuracy: 84.3 -> 82.0 | Hurt |
| replace alpha blending with StochasticSplats at 1 SPP | LangSplat · LERF Waldo-Kitchen | accuracy: 95.5 -> 86.4 | Hurt |
| replace alpha blending with StochasticSplats at 1 SPP | LangSplat · LERF Teatime, Ramen, Figurines | accuracy: 88.1 -> 88.1 | No clear effect |
| add temporal anti-aliasing to 1 SPP rendering (ablation) | StochasticSplats · 1 SPP rendering | noise | Improved |
| lower rendering resolution of sorted alpha blending (ablation) | 3DGS-CUDA · render time at reduced resolution | render time | No clear effect |
Closest papers by meaning
- Sort-free Gaussian Splatting via Weighted Sum Rendering (ICLR 2025 · similarity 0.92)
- FlashGS: Efficient 3D Gaussian Splatting for Large-scale and High-resolution Rendering (CVPR 2025 · similarity 0.92)
- 3D Convex Splatting: Radiance Field Rendering with 3D Smooth Convexes (CVPR 2025 · similarity 0.92)
- Speedy-Splat: Fast 3D Gaussian Splatting with Sparse Pixels and Sparse Primitives (CVPR 2025 · similarity 0.91)
- Volumetrically Consistent 3D Gaussian Rasterization (CVPR 2025 · similarity 0.91)
- DiffGS: Functional Gaussian Splatting Diffusion (NeurIPS 2024 · similarity 0.91)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
