Dev

TIRF — 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
incidence_angle -0.36 – 0.54 deg
penetration_depth -24.0 – 36.0 nm
refractive_index 1.509 – 1.524

Dev Tier Leaderboard

# Method Score PSNR SSIM Consistency Trust Source
1 DeconvFormer + gradient 0.755 31.98 0.942 0.79 ✓ Certified Chen et al., CVPR 2024
2 Restormer+ + gradient 0.754 30.57 0.925 0.89 ✓ Certified Zamir et al., ICCV 2024
3 DiffDeconv + gradient 0.726 28.86 0.898 0.89 ✓ Certified Huang et al., NeurIPS 2024
4 Restormer + gradient 0.724 29.43 0.908 0.83 ✓ Certified Zamir et al., CVPR 2022
5 ScoreMicro + gradient 0.716 28.85 0.898 0.84 ✓ Certified Wei et al., ECCV 2025
6 PnP-FISTA + gradient 0.676 26.94 0.857 0.82 ✓ Certified Bai et al., 2020
7 TV-Deconvolution + gradient 0.675 26.78 0.853 0.83 ✓ Certified Rudin et al., Phys. A 1992
8 ResUNet + gradient 0.655 26.01 0.832 0.81 ✓ Certified DeCelle et al., Nat. Methods 2021
9 PnP-DnCNN + gradient 0.649 25.66 0.822 0.82 ✓ Certified Zhang et al., IEEE TIP 2017
10 CARE + gradient 0.628 24.28 0.778 0.87 ✓ Certified Weigert et al., Nat. Methods 2018
11 U-Net + gradient 0.627 24.76 0.794 0.81 ✓ Certified Ronneberger et al., MICCAI 2015
12 Wiener Filter + gradient 0.622 23.95 0.767 0.88 ✓ Certified Analytical baseline
13 Richardson-Lucy + gradient 0.576 22.28 0.702 0.86 ✓ Certified Richardson, JOSA 1972 / Lucy, AJ 1974

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 TIRF