Dev

3DGS — Dev Tier

(5 scenes)

Blind evaluation tier — no ground truth available.

What you get

Measurements (y), ideal forward operator (H), and spec ranges only.

How to use

Apply your pipeline from the Public tier. Use consistency as self-check.

What to submit

Reconstructed signals and corrected spec. Scored server-side.

Parameter Specifications

🔒

True spec hidden — estimate parameters from spec ranges below.

Parameter Spec Range Unit
camera_pose -1.2 – 1.8 mm/deg
focal_length -6.0 – 9.0 pixels
point_cloud_init -2.4 – 3.6 mm

Dev Tier Leaderboard

# Method Score PSNR SSIM Consistency Trust Source
1 GaussianShader + gradient 0.730 29.57 0.91 0.85 ✓ Certified Wang et al., ICCV 2024
2 2DGS + gradient 0.707 27.92 0.879 0.88 ✓ Certified Huang et al., CVPR 2024
3 NeRFactor2 + gradient 0.691 27.17 0.862 0.87 ✓ Certified Barron et al., NeurIPS 2024
4 3D-GS++ + gradient 0.681 27.0 0.858 0.84 ✓ Certified Kerbl et al., SIGGRAPH 2024
5 3D-GS + gradient 0.680 27.23 0.864 0.81 ✓ Certified Kerbl et al., SIGGRAPH 2023
6 NeRF + gradient 0.638 24.62 0.79 0.88 ✓ Certified Mildenhall et al., ECCV 2020
7 Photogrammetry + gradient 0.618 23.85 0.763 0.87 ✓ Certified Structure-from-Motion baseline
8 COLMAP+MVS + gradient 0.571 22.69 0.719 0.78 ✓ Certified Schonberger & Frahm, CVPR 2016
9 Instant-NGP + gradient 0.544 21.63 0.674 0.79 ✓ Certified Muller et al., SIGGRAPH 2022
10 Mesh-GS + gradient 0.518 20.71 0.632 0.79 ✓ Certified Li et al., ECCV 2024
11 Mip-NeRF 360 + gradient 0.457 18.11 0.505 0.87 ✓ Certified Barron et al., CVPR 2022

Visible Data Fields

y H_ideal spec_ranges

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%
Back to 3DGS