Research map / 3D vision / NeRF and radiance fields
MS-NeRF: Multi-Space Neural Radiance Fields
Ze-Xin Yin, Pengyi Jiao, Jiaxiong Qiu, Ming‐Ming Cheng, Bo Ren
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 neural feature fields in place of direct RGB-G outputs (ablation) | MS-Mip-NeRF 360 · Synthetic mirror dataset | PSNR: 37.50 -> 38.84 | Improved |
| use the MLP-based feature-field module instead of the hybrid module (ablation) | MS-Mip-NeRF 360 · Synthetic mirror dataset | PSNR: 33.59 -> 38.84 | Improved |
| gate from positional-encoding MLP instead of grid features (ablation) | MS-TensoRF · Synthetic mirror dataset, 10 scenes | PSNR: 32.07 -> 34.72 | Improved |
| add the MS module | Mip-NeRF 360 · Synthetic mirror dataset, circle path, reflective areas | PSNR: 24.20 -> 28.35 | Improved |
| add the MS module | Mip-NeRF 360 · Synthetic mirror dataset, circle path, other areas | PSNR: 37.06 -> 37.65 | No clear effect |
| add the MS module | Mip-NeRF · Realistic Synthetic 360 | PSNR: 30.74 -> 30.81 | No clear effect |
| add the MS module | Mip-NeRF · Real Forward-Facing | PSNR: 25.78 -> 25.59 | No clear effect |
| add the MS module with only 30 input views, mirror-passing path (ablation) | Mip-NeRF 360 · Scene01-05 mirror-passing-through path | PSNR: 21.60 -> 21.58 (reflective), 25.11 -> 23.04 (other) | Mixed |
| add the MS module with 30 input views, circle path (ablation) | Mip-NeRF 360 · Scene01-05 circle path | PSNR: 19.08 -> 28.37 | Improved |
| use feature fields instead of radiance fields in iNGP (ablation) | iNGP · Realistic Synthetic 360 | PSNR: 33.43 -> 33.25 | Hurt |
| add the MS module to NeuS | NeuS · Synthetic mirror dataset, circle path | PSNR: 25.86 -> 28.65 | Improved |
| use MSE loss instead of MAE loss (ablation) | MS-NeRF_B · Scene01-05 circle path | PSNR: 24.31 -> 28.09 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 5 of 5 on most shared results.
- Mip-NeRF 360 — Mip-NeRF 360: Unbounded Anti-Aliased Neural Radiance Fields (CVPR 2022) · better on 7 of 7 shared results
- Ref-NeRF — Ref-NeRF: Structured View-Dependent Appearance for Neural Radiance Fields (CVPR 2022) · better on 2 of 2 shared results
- Mip-NeRF — Mip-NeRF: A Multiscale Representation for Anti-Aliasing Neural Radiance Fields (ICCV 2021) · better on 2 of 2 shared results
- NeRFReN — NeRFReN: Neural Radiance Fields With Reflections (CVPR 2022) · better on 1 of 1 shared result
- NeuS — NeuS: Learning Neural Implicit Surfaces by Volume Rendering for Multi-view Reconstruction (NeurIPS 2021) · better on 1 of 1 shared result
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
- Multi-Space Neural Radiance Fields (CVPR 2023 · similarity 0.98)
- PyNeRF: Pyramidal Neural Radiance Fields (NeurIPS 2023 · similarity 0.93)
- NeRF-MS: Neural Radiance Fields with Multi-Sequence (ICCV 2023 · similarity 0.93)
- Rad-NeRF: Ray-decoupled Training of Neural Radiance Field (NeurIPS 2024 · similarity 0.91)
- A General Implicit Framework for Fast NeRF Composition and Rendering (AAAI 2024 · similarity 0.91)
- Is Vanilla MLP in Neural Radiance Field Enough for Few-shot View Synthesis? (CVPR 2024 · 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.
