Dev

CDI — 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
support -3.6 – 5.4 pixels
saturation -6.0 – 9.0 %
missing_center -3.6 – 5.4 pixels

Dev Tier Leaderboard

# Method Score PSNR SSIM Consistency Trust Source
1 AutoPhase++ + gradient 0.731 30.41 0.923 0.79 ✓ Certified Rivenson et al., ECCV 2024
2 HolographyViT + gradient 0.728 30.35 0.922 0.78 ✓ Certified Wang et al., ICCV 2024
3 PhaseFormer + gradient 0.720 28.4 0.889 0.9 ✓ Certified Tian et al., ICCV 2024
4 LRGS + gradient 0.693 27.49 0.87 0.85 ✓ Certified Choi et al., 2023
5 DiffusionPhase + gradient 0.690 27.63 0.873 0.82 ✓ Certified Song et al., NeurIPS 2024
6 ScorePhase + gradient 0.678 26.26 0.839 0.9 ✓ Certified Wei et al., ECCV 2025
7 PhaseResNet + gradient 0.662 25.97 0.831 0.85 ✓ Certified Baoqing et al., Optica 2023
8 CyclePhase + gradient 0.598 23.24 0.74 0.85 ✓ Certified Ge et al., IEEE Photonics 2023
9 PhaseNet + gradient 0.575 22.1 0.694 0.88 ✓ Certified Rivenson et al., LSA 2018
10 GS/HIO + gradient 0.556 21.69 0.677 0.84 ✓ Certified Fienup, Appl. Opt. 1982
11 prDeep + gradient 0.552 21.2 0.655 0.89 ✓ Certified Metzler et al., ICML 2018
12 Error Reduction + gradient 0.504 20.07 0.602 0.81 ✓ Certified Fienup, J. Opt. Soc. Am. 1982
13 deep-PR + gradient 0.451 18.44 0.522 0.79 ✓ Certified Asif et al., ICCP 2017
14 Gerchberg-Saxton + gradient 0.428 17.12 0.456 0.87 ✓ Certified Gerchberg & Saxton, Optik 1972

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 CDI