Public

CASSI — Public Tier

(10 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
mask_dx 0.3 – 0.7 0.5 px
mask_dy 0.1 – 0.5 0.3 px
mask_rotation 0.0 – 0.2 0.1 deg
dispersion_slope 1.895 – 2.145 2.02 px/band
dispersion_axis 0.0 – 0.3 0.15 deg

InverseNet Baseline Scores

Method: CPU_baseline — Mismatch parameter: nominal

Scenario I (Ideal)

11.58 dB

SSIM 0.0182

Scenario II (Mismatch)

16.29 dB

SSIM 0.0421

Scenario III (Oracle)

16.36 dB

SSIM 0.0457

Per-scene breakdown (4 scenes)
Scene PSNR I SSIM I PSNR II SSIM II PSNR III SSIM III
scene_00 11.39 0.0173 16.29 0.0414 16.42 0.0455
scene_01 11.60 0.0187 16.26 0.0420 16.34 0.0458
scene_02 11.33 0.0171 16.35 0.0433 16.28 0.0458
scene_03 11.98 0.0197 16.25 0.0416 16.38 0.0458

Public Tier Leaderboard

# Method Score PSNR SSIM Consistency Trust Source
1 SSR-L + gradient 0.877 38.03 0.994 1.0 ✓ Certified PWM benchmark (CVPR 2024)
2 MST-L + gradient 0.794 31.29 0.977 0.95 ✓ Certified PWM benchmark
3 HDNet + gradient 0.707 25.45 0.921 0.92 ✓ Certified PWM benchmark
4 GAP-TV + gradient 0.687 24.21 0.865 1.0 ✓ Certified PWM benchmark
5 PnP-HSICNN + gradient 0.549 20.06 0.675 0.89 ✓ Certified PWM benchmark

Visible Data Fields

y H_ideal spec_ranges x_true true_spec

Dataset

Format: HDF5
Scenes: 10

Scoring Formula

0.4 × PSNR_norm + 0.4 × SSIM + 0.2 × (1 − ‖y − Ĥx̂‖/‖y‖)

PSNR: 40% SSIM: 40% Consistency: 20%
Back to CASSI