Research map / 3D vision / Gaussian splatting
Boost Your NeRF: A Model-Agnostic Mixture of Experts Framework for High Quality and Efficient Rendering
Francesco Di Sario, Riccardo Renzulli, Marco Grangetto, Enzo Tartaglione
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 resolution-based sparse MoE (Top-2) over multi-resolution experts | DVGO · NSVF | PSNR: 35.21 -> 37.59 | Improved |
| add resolution-based sparse MoE (Top-2) | TensoRF · Synthetic-NeRF | PSNR: 32.98 -> 34.09 | Improved |
| add resolution-based sparse MoE (Top-2) | DVGO · LLFF | PSNR: 26.24 -> 26.62 | Improved |
| add resolution-based sparse MoE (Top-1) | Instant-NGP · LLFF | PSNR: 24.97 -> 24.90 | No clear effect |
| add MoE to Instant-NGP | Instant-NGP · Synthetic-NeRF | LPIPS: 0.025 -> 0.043 (Top-2) | Hurt |
| use Top-2 routing instead of full ensemble (ablation) | DVGO MoE · Synthetic-NeRF | GFLOPs: 2206 -> 940 with PSNR 33.79 -> 33.74 | Improved |
| increase k from 2 to 3-5 (ablation) | DVGO MoE · Synthetic-NeRF | PSNR | No clear effect |
| add resolution penalty to the auxiliary loss (geometric progression) (ablation) | DVGO MoE Top-2 · Synthetic-NeRF | PSNR: 33.61 -> 33.74 | Improved |
| add resolution penalty to the auxiliary loss (ablation) | DVGO MoE Top-2 · Synthetic-NeRF | active parameters (x10^6): 56 -> 39 | Improved |
| increase gate resolution (ablation) | MoE gate · Synthetic-NeRF (DVGO) | PSNR | Hurt |
| use proposed grid+MLP gate instead of Switch-NeRF gate (ablation) | MoE gate · Synthetic-NeRF (DVGO) | PSNR and cost | Improved |
| train MoE end-to-end without pretraining experts independently (ablation) | MoE · general | reconstruction quality | Hurt |
Closest papers by meaning
- Enhancing NeRF akin to Enhancing LLMs: Generalizable NeRF Transformer with Mixture-of-View-Experts (ICCV 2023 · similarity 0.89)
- OPONeRF: One-Point-One NeRF for Robust Few-shot Rendering (IJCV 2025 · similarity 0.89)
- PaReNeRF: Toward Fast Large-scale Dynamic NeRF with Patch-based Reference (CVPR 2024 · similarity 0.87)
- DiSR-NeRF: Diffusion-Guided View-Consistent Super-Resolution NeRF (CVPR 2024 · similarity 0.87)
- NeRF On-the-go: Exploiting Uncertainty for Distractor-free NeRFs in the Wild (CVPR 2024 · similarity 0.87)
- NeRF Director: Revisiting View Selection in Neural Volume Rendering (CVPR 2024 · similarity 0.86)
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.
