Research map / 3D vision / Gaussian splatting
TurboGS: Accelerating 3D Gaussian Splatting via Error-Guided Sparse Pixel Sampling and Optimization
Zheng Dong, Daifei Qiu, Pinxuan Dai, Ke Xu, Jiamin Xu, Lili He, Rynson W. H. Lau, Weiwei Xu
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 |
|---|---|---|---|
| add error-guided sparse pixel sampling (Err-SPS) (ablation) | 3DGS · Mip-NeRF 360 | training time (s): 891 -> 85 | Improved |
| add error-guided sparse pixel sampling (Err-SPS) (ablation) | 3DGS · Mip-NeRF 360 | PSNR: 27.55 -> 26.58 | Hurt |
| add geometry-aware view sampling (Geo-VS) (ablation) | 3DGS + Err-SPS · Mip-NeRF 360 | PSNR: 26.58 -> 26.73 | Improved |
| add sparse NCC loss (ablation) | 3DGS + Err-SPS + Geo-VS · Mip-NeRF 360 | PSNR: 26.73 -> 27.14 | Improved |
| add error-driven density control (Err-DP) (ablation) | 3DGS + Err-SPS + Geo-VS + sparse NCC · Mip-NeRF 360 | PSNR: 27.14 -> 27.57 | Improved |
| use error-guided sparse pixel sampling instead of random pixel sampling (ablation) | TurboGS · Mip-NeRF 360 | PSNR: 27.10 -> 27.57 | Improved |
| use geometry-aware view sampling instead of random view sampling (ablation) | TurboGS · Mip-NeRF 360 | PSNR: 27.26 -> 27.57 | Improved |
| use error-driven density control instead of FastGS-style density control (ablation) | TurboGS · Mip-NeRF 360 | training time (s): 91 -> 77 | Improved |
| add sparse NCC to l1 loss (ablation) | TurboGS · Mip-NeRF 360 | PSNR: 27.09 -> 27.57 | Improved |
| use sparse NCC instead of sparse SSIM (ablation) | TurboGS · Mip-NeRF 360 | LPIPS: 0.287 -> 0.256 | Improved |
| use moment-damped LM instead of sparse Adam (ablation) | TurboGS · Mip-NeRF 360 | PSNR: 27.52 -> 27.57 | Improved |
| use adaptive view budget 10 to 3 instead of fixed K=1 (ablation) | TurboGS · Mip-NeRF 360 | PSNR: 26.50 -> 27.57 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- Speedy-Splat — Speedy-Splat: Fast 3D Gaussian Splatting with Sparse Pixels and Sparse Primitives (CVPR 2025) · better on 2 of 2 shared results
- Mini-Splatting — Mini-Splatting: Representing Scenes with a Constrained Number of Gaussians (ECCV 2024) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- FastGS: Training 3D Gaussian Splatting in 100 Seconds (CVPR 2026 · similarity 0.94)
- 3DGS-LM: Faster Gaussian-Splatting Optimization with Levenberg-Marquardt (ICCV 2025 · similarity 0.94)
- Speedy-Splat: Fast 3D Gaussian Splatting with Sparse Pixels and Sparse Primitives (CVPR 2025 · similarity 0.93)
- DashGaussian: Optimizing 3D Gaussian Splatting in 200 Seconds (CVPR 2025 · similarity 0.92)
- AdpSplit: Error-Driven Adaptive Splitting for Faster Geometry Discovery in 3D Gaussian Splatting (NeurIPS 2026 · similarity 0.92)
- Steepest Descent Density Control for Compact 3D Gaussian Splatting (CVPR 2025 · similarity 0.92)
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.
