Research map / 3D vision / Gaussian splatting
Scene Experts: Specializing in 3D Gaussian Splatting with Adaptive Decomposition
Xiaowen Fu, Yang Zhang, Yuhan Tang, Huazhong Zhang, Tianxing Zhao, Yuhang Guo, Yu Huang, Jinbao Wang
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 |
|---|---|---|---|
| replace single MLP decoder with sparse MoE of 5 experts, Top-2 (ablation) | MoE-GS · DeepBlending | PSNR: 30.37 -> 30.64 | Improved |
| replace single MLP decoder with MoE (ablation) | MoE-GS · Tanks&Temples | PSNR: 24.30 -> 24.61 | Improved |
| add dense-to-sparse curriculum learning (ablation) | MoE-GS · DeepBlending | PSNR: 30.47 -> 30.64 | Improved |
| add load balancing loss (ablation) | MoE-GS · DeepBlending | PSNR: 30.53 -> 30.64 | Improved |
| add load balancing loss (ablation) | MoE-GS · Quebec (BungeeNeRF), covariance expert | anchor share of busiest expert: 73.1% -> 33.7% | Improved |
| add opacity-aware regularization loss (ablation) | MoE-GS · DeepBlending | PSNR: 30.52 -> 30.64 | Improved |
| add opacity-aware regularization loss (ablation) | MoE-GS · DeepBlending | model size (MB): 58.6 -> 55.4 | Improved |
| add opacity-aware regularization to Scaffold-GS (ablation) | MoE-GS · MipNeRF360 | Gaussian activation ratio: 36.9 -> 63.1 | Improved |
| increase number of experts from 2 to 5 (ablation) | MoE-GS · DeepBlending | PSNR | Improved |
| use Top-1 routing instead of Top-2 (ablation) | MoE-GS · DeepBlending | PSNR | Hurt |
| use dense routing instead of Top-2 (ablation) | MoE-GS · DeepBlending | training time | Mixed |
| use anchor positions as router input (ablation) | MoE-GS · DeepBlending | PSNR | No clear effect |
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.
- Scaffold-GS — Scaffold-GS: Structured 3D Gaussians for View-Adaptive Rendering (CVPR 2024) · better on 7 of 7 shared results
- SAGS — SAGS: Structure-Aware 3D Gaussian Splatting (ECCV 2024) · better on 4 of 4 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
- MoE-GS: Mixture of Experts for Dynamic Gaussian Splatting (ICLR 2026 · similarity 0.93)
- CAG-GS: Consistent Anchor Guided Gaussian Splatting for Large-scale Scene Rendering (AAAI 2026 · similarity 0.91)
- SfM-Free 3D Gaussian Splatting via Hierarchical Training (CVPR 2025 · similarity 0.90)
- FlexGS: Train Once, Deploy Everywhere with Many-in-One Flexible 3D Gaussian Splatting (CVPR 2025 · similarity 0.90)
- Speedy-Splat: Fast 3D Gaussian Splatting with Sparse Pixels and Sparse Primitives (CVPR 2025 · similarity 0.90)
- Generative Sparse-View Gaussian Splatting (CVPR 2025 · similarity 0.90)
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.
