Research map / 3D vision / NeRF and radiance fields
Invertible Neural Warp for NeRF
Shin-Fang Chng, Ravi Garg, Hemanth Saratchandran, Simon Lucey
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 |
|---|---|---|---|
| enforce approximate invertibility with an auxiliary backward network (ablation) | MLP ray warp · 2D planar image alignment | success rate: 0.00 -> 0.65 | Improved |
| use an explicitly invertible INN instead of an implicit-invertible MLP (ablation) | MLP ray warp · 2D planar image alignment | success rate: 0.65 -> 0.75 | Improved |
| use explicit INN warp instead of SE(3) pose parameterization | BARF · 2D planar image alignment | corner error (px): 29.63 -> 4.70 | Improved |
| naive MLP overparameterization without invertibility (ablation) | ray warp · 2D planar image alignment | success rate: 0.30 (BARF) -> 0.00 (Naive) | Hurt |
| replace SE(3) pose with INN overparameterized warp on LLFF | BARF · LLFF | rotation error (deg): 0.90 -> 0.31 | Improved |
| INN versus L2G after test-time pose refinement | LLFF novel view synthesis · LLFF | PSNR: 24.35 (L2G) vs 24.28 (INN) | No clear effect |
| replace SE(3) pose with INN warp on noisy DTU initialization | BARF · DTU | rotation error (deg): 2.52 -> 1.17 | Improved |
| INN versus L2G on DTU geometry quality | pose-NeRF joint optimization · DTU | Chamfer distance: 6.53 -> 4.89 | Improved |
| single shared INN with per-frame latent code instead of one network per frame (ablation) | INN warp · 2D planar image alignment | convergence | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- BARF — BARF: Bundle-Adjusting Neural Radiance Fields (ICCV 2021) · better on 11 of 11 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
- A-NeRF: Articulated Neural Radiance Fields for Learning Human Shape, Appearance, and Pose (NeurIPS 2021 · similarity 0.89)
- NoPe-NeRF: Optimising Neural Radiance Field With No Pose Prior (CVPR 2023 · similarity 0.89)
- Pose-Free Neural Radiance Fields via Implicit Pose Regularization (ICCV 2023 · similarity 0.89)
- Joint Optimization of Neural Radiance Fields and Continuous Camera Motion from a Monocular Video (CVPR 2025 · similarity 0.88)
- ICON: Incremental CONfidence for Joint Pose and Radiance Field Optimization (CVPR 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.
