Research map / 3D vision / NeRF and radiance fields
LU-NeRF: Scene and Pose Estimation by Synchronizing Local Unposed NeRFs
Zezhou Cheng, Carlos Esteves, Varun Jampani, Abhishek Kar, Subhransu Maji, Ameesh Makadia
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 LU-NeRF-2 stage to resolve mirror-symmetry ambiguity (ablation) | LU-NeRF pipeline · Blender unordered, Hotdog | mean rotation error (deg): 138.9 -> 0.23 | Improved |
| add LU-NeRF-2 stage to resolve mirror-symmetry ambiguity (ablation) | LU-NeRF pipeline · Blender unordered, Chair | mean rotation error (deg): 39.14 -> 4.24 | Improved |
| add LU-NeRF-2 stage to resolve mirror-symmetry ambiguity (ablation) | LU-NeRF pipeline · Blender unordered, Lego | mean rotation error (deg): 0.48 -> 0.07 | Improved |
| use LU-NeRF instead of GNeRF 180 (2DOF) for pose estimation | unposed NeRF · Blender unordered | rotation error (deg): 8.77 -> 0.09 (Lego); 36.74 -> 0.24 (Hotdog) | Improved |
| use LU-NeRF instead of GNeRF 180 on Drums scene | unposed NeRF · Blender unordered, Drums | rotation error (deg): 9.01 -> 12.39 | Hurt |
| widen GNeRF elevation prior from 90 to 180 degrees | GNeRF · Blender unordered, Chair | PSNR: 31.30 -> 21.27 | Hurt |
| use LU-NeRF with ordered image collection graph | unposed NeRF · Blender ordered collections | mean rotation error (deg): 8.91 -> 0.76 (vs GNeRF 90) | Improved |
| use LU-NeRF versus GNeRF 90 for view synthesis on ordered Lego | unposed NeRF · Blender ordered, Lego | PSNR: 22.83 -> 15.90 | Hurt |
| lower image resolution for COLMAP from 800x800 to 400x400 | COLMAP · Blender, Mic | images registered out of 100: 80 -> 15 | Hurt |
| use LU-NeRF instead of COLMAP variants on textureless Objectron scenes | pose estimation · Objectron | rotation error (deg): Chair 6.7 -> 2.6; Bike 1.1 -> 15.6 (vs COLMAP-LoFTR) | Mixed |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- GNeRF 90 — GNeRF: GAN-Based Neural Radiance Field Without Posed Camera (ICCV 2021) · better on 3 of 6 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
- UP-NeRF: Unconstrained Pose Prior-Free Neural Radiance Field (NeurIPS 2023 · similarity 0.90)
- Pose-Free Neural Radiance Fields via Implicit Pose Regularization (ICCV 2023 · similarity 0.90)
- ICON: Incremental CONfidence for Joint Pose and Radiance Field Optimization (CVPR 2024 · similarity 0.88)
- LuSh-NeRF: Lighting up and Sharpening NeRFs for Low-light Scenes (NeurIPS 2024 · similarity 0.87)
- Joint Optimization of Neural Radiance Fields and Continuous Camera Motion from a Monocular Video (CVPR 2025 · similarity 0.87)
- CF-NeRF: Camera Parameter Free Neural Radiance Fields with Incremental Learning (AAAI 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.
