Public
NeRF — Public Tier
(5 scenes)Full-access development tier with all data visible.
What you get
Measurements (y), ideal forward operator (H), spec ranges, ground truth (x_true), and true mismatch spec.
How to use
Load HDF5 → compare reconstruction vs x_true → check consistency → iterate.
What to submit
Reconstructed signals (x_hat) and corrected spec as HDF5.
Parameter Specifications
✓
True spec visible — use these exact values for Scenario III oracle reconstruction.
| Parameter | Spec Range | True Value | Unit |
|---|---|---|---|
| camera_pose | -1.0 – 2.0 | 0.5 | mm/deg |
| focal_length | -5.0 – 10.0 | 2.5 | pixels |
| distortion | -0.01 – 0.02 | 0.005 | |
| exposure | -10.0 – 20.0 | 5.0 | % |
InverseNet Baseline Scores
Method: CPU_baseline — Mismatch parameter: nominal
Scenario I (Ideal)
21.37 dB
SSIM 0.8091
Scenario II (Mismatch)
33.72 dB
SSIM 0.9280
Scenario III (Oracle)
46.41 dB
SSIM 0.9898
Per-scene breakdown (4 scenes)
| Scene | PSNR I | SSIM I | PSNR II | SSIM II | PSNR III | SSIM III |
|---|---|---|---|---|---|---|
| scene_00 | 21.33 | 0.8107 | 33.61 | 0.9287 | 46.07 | 0.9896 |
| scene_01 | 21.54 | 0.8154 | 33.80 | 0.9299 | 46.27 | 0.9902 |
| scene_02 | 20.21 | 0.7912 | 32.90 | 0.9251 | 46.42 | 0.9898 |
| scene_03 | 22.40 | 0.8190 | 34.57 | 0.9282 | 46.88 | 0.9898 |
Public Tier Leaderboard
| # | Method | Score | PSNR | SSIM | Consistency | Trust | Source |
|---|---|---|---|---|---|---|---|
| 1 | NeRFactor2 + gradient | 0.818 | 34.58 | 0.965 | 0.93 | ✓ Certified | Barron et al., NeurIPS 2024 |
| 2 | GaussianShader + gradient | 0.783 | 32.45 | 0.947 | 0.9 | ✓ Certified | Wang et al., ICCV 2024 |
| 3 | 3D-GS++ + gradient | 0.777 | 32.27 | 0.946 | 0.88 | ✓ Certified | Kerbl et al., SIGGRAPH 2024 |
| 4 | 3D-GS + gradient | 0.761 | 31.19 | 0.933 | 0.88 | ✓ Certified | Kerbl et al., SIGGRAPH 2023 |
| 5 | Instant-NGP + gradient | 0.748 | 29.58 | 0.91 | 0.94 | ✓ Certified | Muller et al., SIGGRAPH 2022 |
| 6 | 2DGS + gradient | 0.733 | 29.49 | 0.909 | 0.87 | ✓ Certified | Huang et al., CVPR 2024 |
| 7 | NeRF + gradient | 0.726 | 28.99 | 0.9 | 0.88 | ✓ Certified | Mildenhall et al., ECCV 2020 |
| 8 | Mip-NeRF 360 + gradient | 0.717 | 27.69 | 0.874 | 0.95 | ✓ Certified | Barron et al., CVPR 2022 |
| 9 | Mesh-GS + gradient | 0.713 | 28.26 | 0.886 | 0.88 | ✓ Certified | Li et al., ECCV 2024 |
| 10 | COLMAP+MVS + gradient | 0.653 | 24.64 | 0.791 | 0.95 | ✓ Certified | Schonberger & Frahm, CVPR 2016 |
| 11 | Photogrammetry + gradient | 0.631 | 24.41 | 0.783 | 0.87 | ✓ Certified | Structure-from-Motion baseline |
Visible Data Fields
y
H_ideal
spec_ranges
x_true
true_spec
Dataset
Format: HDF5
Scenes: 5
Scoring Formula
0.4 × PSNR_norm + 0.4 × SSIM + 0.2 × (1 − ‖y − Ĥx̂‖/‖y‖)
PSNR: 40%
SSIM: 40%
Consistency: 20%