Research map / 3D vision / NeRF and radiance fields
Learning Robust Multi-scale Representation for Neural Radiance Fields from Unposed Images
Nishant Jain, Suryansh Kumar, Luc Van Gool
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 |
|---|---|---|---|
| Naively combine Mip-NeRF with BARF pose refinement instead of using RM-NeRF | Mip-NeRF · Multi-scale Blender Lego (rotation noise) | PSNR: 27.01 -> 11.67 | Hurt |
| Add joint GNN motion averaging pose refinement to multi-scale NeRF (RM-NeRF vs Mip-NeRF) | Mip-NeRF · Multi-scale Blender Lego (rotation noise) | PSNR: 21.52 -> 27.01 | Improved |
| Initialize RM-NeRF (w/o pose) from the same noisy poses as RM-NeRF | RM-NeRF (w/o pose) · Multi-scale Blender Lego (rotation noise) | PSNR: 26.34 -> 27.91 | Improved |
| Add translation noise to the perturbed poses | RM-NeRF · Multi-scale Blender Lego | PSNR: 27.01 -> 26.10 | Hurt |
| Double the pose noise standard deviation | RM-NeRF · Multi-scale Blender Lego (increased noise) | PSNR: 26.10 -> 23.21 | Hurt |
| Use random pose initialization (w/o pose) instead of noisy COLMAP-style poses under increased noise | RM-NeRF · Multi-scale Blender Lego (increased noise) | PSNR: 23.21 -> 26.34 | Improved |
| Replace RM-NeRF with RM-NeRF (w/o pose) when both start from COLMAP poses | RM-NeRF · Tanks and Temples Truck (same initialization as RM-NeRF) | PSNR: 26.12 -> 26.93 | Improved |
| Replace RM-NeRF with RM-NeRF (E2E) when both start from COLMAP poses | RM-NeRF · Tanks and Temples Truck (same initialization as RM-NeRF) | PSNR: 26.12 -> 26.57 | Improved |
| Add robust motion averaging and depth priors to random-pose training (RM-NeRF w/o pose vs NoPe-NeRF) | NoPe-NeRF · Tanks and Temples Truck (COLMAP intrinsics, random poses) | PSNR: 25.28 -> 25.81 | Improved |
| Learn intrinsics jointly (RM-NeRF E2E) instead of NeRF-- without COLMAP | NeRF-- · Tanks and Temples Truck (no COLMAP poses or intrinsics) | PSNR: 24.98 -> 25.12 | Improved |
| Add RM-NeRF refinement to COLMAP-pose Mip-NeRF | Mip-NeRF · Tanks and Temples Truck (COLMAP poses and intrinsics) | PSNR: 25.37 -> 26.12 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- Mip-NeRF — Mip-NeRF: A Multiscale Representation for Anti-Aliasing Neural Radiance Fields (ICCV 2021) · better on 7 of 7 shared results
- BARF — BARF: Bundle-Adjusting Neural Radiance Fields (ICCV 2021) · better on 3 of 3 shared results
- NoPe-NeRF — NoPe-NeRF: Optimising Neural Radiance Field With No Pose Prior (CVPR 2023) · 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
- Joint Optimization of Neural Radiance Fields and Continuous Camera Motion from a Monocular Video (CVPR 2025 · similarity 0.89)
- Non-Rigid Neural Radiance Fields: Reconstruction and Novel View Synthesis of a Dynamic Scene From Monocular Video (ICCV 2021 · similarity 0.89)
- Progressively Optimized Local Radiance Fields for Robust View Synthesis (CVPR 2023 · similarity 0.88)
- AltNeRF: Learning Robust Neural Radiance Field via Alternating Depth-Pose Optimization (AAAI 2024 · similarity 0.88)
- Few-Shot Neural Radiance Fields under Unconstrained Illumination (AAAI 2024 · similarity 0.87)
- CF-NeRF: Camera Parameter Free Neural Radiance Fields with Incremental Learning (AAAI 2024 · similarity 0.87)
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.
