Stellar Coronagraphy

Stellar Coronagraphy

Join the Competition Contribute to Benchmark

Standard reconstruction benchmark — forward model perfectly known, no calibration needed. Score = 0.5 × clip((PSNR−15)/30, 0, 1) + 0.5 × SSIM

# Method Score PSNR (dB) SSIM Source
🥇 DiffusionCoron 0.876 38.9 0.955 ✓ Certified Lim 2024
🥈 CoronFormer 0.831 36.8 0.935 ✓ Certified Gebhard 2022
🥉 SpeckleLearn 0.780 34.5 0.910 ✓ Certified Yip 2020
4 CNN-Coronagraph 0.724 32.1 0.878 ✓ Certified Gonzalez 2018
5 ANDROMEDA 0.649 28.8 0.838 ✓ Certified Cantalloube 2015
6 KLIP 0.616 27.5 0.815 ✓ Certified Soummer 2012
7 PCA-ADI 0.582 26.2 0.791 ✓ Certified Amara 2012
8 LOCI 0.544 24.8 0.762 ✓ Certified Lafrenière 2007
9 ADI 0.485 22.5 0.721 ✓ Certified Marois 2006

Dataset: PWM Benchmark (9 algorithms)

Blind Reconstruction Challenge — forward model has unknown mismatch, must calibrate from data. Score = 0.4 × PSNR_norm + 0.4 × SSIM + 0.2 × (1 − ‖y − Ĥx̂‖/‖y‖)

# Method Overall Score Public
PSNR / SSIM
 Dev
PSNR / SSIM
 Hidden
PSNR / SSIM
 Trust Source
🥇 DiffusionCoron + gradient 0.768
0.835
36.95 dB / 0.978
0.746
30.34 dB / 0.922
0.722
29.96 dB / 0.916
✓ Certified Lim et al., ApJ 2024
🥈 CoronFormer + gradient 0.728
0.808
34.57 dB / 0.965
0.731
29.05 dB / 0.901
0.646
25.96 dB / 0.831
✓ Certified Gebhard et al., A&A 2022
🥉 SpeckleLearn + gradient 0.685
0.775
31.87 dB / 0.941
0.679
26.89 dB / 0.855
0.600
23.3 dB / 0.743
✓ Certified Yip et al., AJ 2020
4 CNN-Coronagraph + gradient 0.641
0.742
30.1 dB / 0.918
0.616
23.8 dB / 0.761
0.565
21.97 dB / 0.689
✓ Certified Gonzalez et al., AJ 2018
5 KLIP + gradient 0.634
0.682
26.15 dB / 0.836
0.625
24.14 dB / 0.773
0.596
23.12 dB / 0.736
✓ Certified Soummer et al., ApJ 2012
6 PCA-ADI + gradient 0.592
0.627
24.24 dB / 0.777
0.574
22.82 dB / 0.724
0.575
22.69 dB / 0.719
✓ Certified Amara & Quanz, MNRAS 2012
7 LOCI + gradient 0.584
0.619
23.41 dB / 0.747
0.591
23.01 dB / 0.731
0.542
20.92 dB / 0.642
✓ Certified Lafrenière et al., ApJ 2007
8 ANDROMEDA + gradient 0.565
0.672
25.87 dB / 0.828
0.543
21.54 dB / 0.670
0.479
18.97 dB / 0.548
✓ Certified Cantalloube et al., A&A 2015
9 ADI + gradient 0.491
0.514
19.81 dB / 0.590
0.492
19.82 dB / 0.590
0.468
18.59 dB / 0.529
✓ Certified Marois et al., ApJ 2006

Complete score requires all 3 tiers (Public + Dev + Hidden).

Join the competition →
Scoring: 0.4 × PSNR_norm + 0.4 × SSIM + 0.2 × (1 − ‖y − Ĥx̂‖/‖y‖) PSNR 40% · SSIM 40% · Consistency 20%
Public 3 scenes

Full-access development tier with all data visible.

What you get & how to use

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.

Public Leaderboard
# Method Score PSNR SSIM
1 DiffusionCoron + gradient 0.835 36.95 0.978
2 CoronFormer + gradient 0.808 34.57 0.965
3 SpeckleLearn + gradient 0.775 31.87 0.941
4 CNN-Coronagraph + gradient 0.742 30.1 0.918
5 KLIP + gradient 0.682 26.15 0.836
6 ANDROMEDA + gradient 0.672 25.87 0.828
7 PCA-ADI + gradient 0.627 24.24 0.777
8 LOCI + gradient 0.619 23.41 0.747
9 ADI + gradient 0.514 19.81 0.59
Spec Ranges (4 parameters)
Parameter Min Max Unit
coronagraph_mask_centering -0.02 0.04 lambda/D
wavefront_error_(wfe) -20.0 40.0 -
stellar_leakage -0.199997 0.4 contrast
speckle_lifetime -20.0 40.0 s
View Details →
Dev 3 scenes

Blind evaluation tier — no ground truth available.

What you get & how to use

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.

Dev Leaderboard
# Method Score PSNR SSIM
1 DiffusionCoron + gradient 0.746 30.34 0.922
2 CoronFormer + gradient 0.731 29.05 0.901
3 SpeckleLearn + gradient 0.679 26.89 0.855
4 KLIP + gradient 0.625 24.14 0.773
5 CNN-Coronagraph + gradient 0.616 23.8 0.761
6 LOCI + gradient 0.591 23.01 0.731
7 PCA-ADI + gradient 0.574 22.82 0.724
8 ANDROMEDA + gradient 0.543 21.54 0.67
9 ADI + gradient 0.492 19.82 0.59
Spec Ranges (4 parameters)
Parameter Min Max Unit
coronagraph_mask_centering -0.024 0.036 lambda/D
wavefront_error_(wfe) -24.0 36.0 -
stellar_leakage -0.239998 0.36 contrast
speckle_lifetime -24.0 36.0 s
Submit Reconstruction View Details →
Hidden 3 scenes

Fully blind server-side evaluation — no data download.

What you get & how to use

What you get: No data downloadable. Algorithm runs server-side on hidden measurements.

How to use: Package algorithm as Docker container / Python script. Submit via link.

What to submit: Containerized algorithm accepting y + H, outputting x_hat + corrected spec.

Hidden Leaderboard
# Method Score PSNR SSIM
1 DiffusionCoron + gradient 0.722 29.96 0.916
2 CoronFormer + gradient 0.646 25.96 0.831
3 SpeckleLearn + gradient 0.600 23.3 0.743
4 KLIP + gradient 0.596 23.12 0.736
5 PCA-ADI + gradient 0.575 22.69 0.719
6 CNN-Coronagraph + gradient 0.565 21.97 0.689
7 LOCI + gradient 0.542 20.92 0.642
8 ANDROMEDA + gradient 0.479 18.97 0.548
9 ADI + gradient 0.468 18.59 0.529
Spec Ranges (4 parameters)
Parameter Min Max Unit
coronagraph_mask_centering -0.014 0.046 lambda/D
wavefront_error_(wfe) -14.0 46.0 -
stellar_leakage -0.139998 0.459998 contrast
speckle_lifetime -14.0 46.0 s
Submit Algorithm View Details →

Blind Reconstruction Challenge

Challenge

Given measurements with unknown mismatch and spec ranges (not exact params), reconstruct the original signal. A method must be evaluated on all three tiers for a complete score. Scored on a composite metric: 0.4 × PSNR_norm + 0.4 × SSIM + 0.2 × (1 − ‖y − Ĥx̂‖/‖y‖).

Input

Measurements y, ideal forward model H, spec ranges

Output

Reconstructed signal x̂

Spec DAG — Forward Model Pipeline

M → P → D

M Modulation
P Propagation
D Detector

Mismatch Parameters

Symbol Parameter Description Nominal Perturbed
c_m coronagraph_mask_centering Coronagraph mask centering (lambda/D) 0.0 0.02
w_e wavefront_error_(wfe) Wavefront error (WFE) (-) 0.0 20.0
s_l stellar_leakage Stellar leakage (contrast) 1e-06 0.2
s_l speckle_lifetime Speckle lifetime (s) 0.0 20.0

Credits System

40%
Platform Profit Pool
Revenue allocated to benchmark rewards
30%
Winner Share
Top algorithm receives from pool
$100
Min Withdrawal
Minimum payout threshold
Spec Primitives Reference (11 primitives)
P Propagation

Free-space or medium propagation kernel (Fresnel, Rayleigh-Sommerfeld).

M Mask / Modulation

Spatial or spatio-temporal amplitude modulation (coded aperture, SLM pattern).

Π Projection

Geometric projection operator (Radon transform, fan-beam, cone-beam).

F Fourier Sampling

Sampling in the Fourier / k-space domain (MRI, ptychography).

C Convolution

Shift-invariant convolution with a point-spread function (PSF).

Σ Summation / Integration

Summation along a physical dimension (spectral, temporal, angular).

D Detector

Sensor readout with gain g and noise model η (Gaussian, Poisson, mixed).

S Structured Illumination

Patterned illumination (block, Hadamard, random) applied to the scene.

W Wavelength Dispersion

Spectral dispersion element (prism, grating) with shift α and aperture a.

R Rotation / Motion

Sample or gantry rotation (CT, electron tomography).

Λ Wavelength Selection

Spectral filter or monochromator selecting a wavelength band.