Research map / 3D vision / NeRF and radiance fields
SPARF: Neural Radiance Fields From Sparse and Noisy Poses
Prune Truong, Marie-Julie Rakotosaona, Fabian Manhardt, Federico Tombari
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 multi-view correspondence supervision to sparse-view NeRF (ablation) | SPARF on DTU with fixed ground-truth poses · DTU, 3 views | PSNR: 16.27 -> 18.13 | Improved |
| Add depth consistency loss to multi-view correspondence loss (ablation) | SPARF on DTU with fixed ground-truth poses · DTU, 3 views | PSNR: 18.13 -> 18.30 | Improved |
| Use depth consistency loss alone, without multi-view correspondence loss (ablation) | SPARF on DTU with fixed ground-truth poses · DTU, 3 views | PSNR: 16.27 -> 15.86 | Hurt |
| Use coarse-to-fine positional encoding instead of full high-frequency encoding (ablation) | NeRF on DTU with fixed ground-truth poses · DTU, 3 views | PSNR: 8.41 -> 16.27 | Improved |
| Remove positional encoding (ablation) | NeRF on DTU with fixed ground-truth poses · DTU, 3 views | PSNR: 8.41 -> 16.11 | Improved |
| Use a smaller MLP (ablation) | NeRF on DTU with fixed ground-truth poses · DTU, 3 views | PSNR: 8.41 -> 9.03 | No clear effect |
| Replace per-image ground-truth depth supervision with ground-truth 3D-point supervision (ablation) | Joint pose-NeRF training on DTU with noisy poses · DTU, 3 views | rotation error: 7.3 -> 0.4 | Improved |
| Add multi-view correspondence loss to photometric training (ablation) | Joint pose-NeRF training on DTU with noisy poses · DTU, 3 views | rotation error: 10.3 -> 1.85 | Improved |
| Add a mask loss to photometric training (ablation) | Joint pose-NeRF training on DTU with noisy poses · DTU, 3 views | rotation error: 10.3 -> 13.2 | Hurt |
| Add multi-view correspondence and jointly refine poses and NeRF | SPARF · DTU, 3 views, noisy poses | PSNR: 12.04 -> 17.74 | Improved |
| Use SPARF rather than RegBARF for sparse pose refinement | NeRF pose refinement · LLFF, 3 views, identity poses | PSNR: 18.57 -> 19.58 | Improved |
| Use SPARF rather than DS-NeRF for sparse pose refinement | NeRF pose refinement · Replica, 3 views, COLMAP poses | PSNR: 24.75 -> 26.98 | Improved |
Later papers that compared with this method
9 later accepted papers report SPARF as a baseline in their results tables; 8 of them report a better number on most of the results they share with it.
- PMNI: Pose-free Multi-view Normal Integration for Reflective and Textureless Surface Reconstruction (CVPR 2025 · better on 2 of 2 shared results)
- Sparfels: Fast Reconstruction from Sparse Unposed Imagery (ICCV 2025 · better on 1 of 1 shared result)
- TrackNeRF: Bundle Adjusting NeRF from Sparse and Noisy Views via Feature Tracks (ECCV 2024 · better on 9 of 11 shared results)
- CRAYM: Neural Field Optimization via Camera RAY Matching (NeurIPS 2024 · better on 3 of 4 shared results)
- CVT-xRF: Contrastive In-Voxel Transformer for 3D Consistent Radiance Fields from Sparse Inputs (CVPR 2024 · better on 3 of 3 shared results)
- ExtraNeRF: Visibility-Aware View Extrapolation of Neural Radiance Fields with Diffusion Models (CVPR 2024 · better on 2 of 3 shared results)
- PORF: POSE RESIDUAL FIELD FOR ACCURATE NEURAL SURFACE RECONSTRUCTION (ICLR 2024 · better on 2 of 2 shared results)
- Revisit Self-supervision with Local Structure-from-Motion (ECCV 2024 · better on 1 of 2 shared results)
1 more not listed.
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
- Neural Radiance Fields From Sparse RGB-D Images for High-Quality View Synthesis (TPAMI 2022 · similarity 0.93)
- Joint Optimization of Neural Radiance Fields and Continuous Camera Motion from a Monocular Video (CVPR 2025 · similarity 0.91)
- CF-NeRF: Camera Parameter Free Neural Radiance Fields with Incremental Learning (AAAI 2024 · similarity 0.91)
- MuRF: Multi-Baseline Radiance Fields (CVPR 2024 · similarity 0.90)
- ColNeRF: Collaboration for Generalizable Sparse Input Neural Radiance Field (AAAI 2024 · similarity 0.90)
- Few-Shot Neural Radiance Fields under Unconstrained Illumination (AAAI 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.
