Adaptive Optics (AO) Imaging

Adaptive Optics (AO) Imaging

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
🥇 DiffusionAO 0.807 35.0 0.948 ✓ Certified Diffusion for wavefront, 2024
🥈 AO-ViT 0.784 34.0 0.935 ✓ Certified Vision transformer for AO, 2024
🥉 AO-Transformer 0.760 33.0 0.920 ✓ Certified AO transformer, 2023
4 LIFT-Net 0.723 31.5 0.895 ✓ Certified Orban de Xivry et al., MNRAS 2021
5 WFNet 0.685 30.0 0.870 ✓ Certified Nishizaki et al., Opt. Express 2019
6 PnP-ADMM (WF) 0.600 27.0 0.800 ✓ Certified Venkatakrishnan et al., 2013
7 Fried Estimator 0.500 24.0 0.700 ✓ Certified Fried, JOSA 1977
8 Zernike LS 0.437 22.0 0.640 ✓ Certified Noll, JOSA 1976

Dataset: PWM Benchmark (8 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
🥇 AO-ViT + gradient 0.709
0.767
31.2 dB / 0.933
0.716
28.32 dB / 0.887
0.645
25.09 dB / 0.805
✓ Certified Vision transformer for AO, 2024
🥈 DiffusionAO + gradient 0.674
0.805
33.44 dB / 0.956
0.654
25.11 dB / 0.806
0.562
21.83 dB / 0.683
✓ Certified Score-based diffusion for wavefront reconstruction, 2024
🥉 AO-Transformer + gradient 0.670
0.751
30.14 dB / 0.919
0.679
27.3 dB / 0.865
0.581
22.48 dB / 0.710
✓ Certified Wavefront sensing transformer, 2023
4 LIFT-Net + gradient 0.638
0.729
29.07 dB / 0.901
0.622
24.03 dB / 0.770
0.563
22.7 dB / 0.719
✓ Certified Orban de Xivry et al., MNRAS 2021
5 WFNet + gradient 0.632
0.698
27.15 dB / 0.862
0.628
24.59 dB / 0.789
0.571
22.63 dB / 0.716
✓ Certified Nishizaki et al., Opt. Express 2019
6 PnP-ADMM (WF) + gradient 0.554
0.640
24.61 dB / 0.790
0.533
21.1 dB / 0.650
0.489
19.52 dB / 0.575
✓ Certified Venkatakrishnan et al., 2013
7 Fried Estimator + gradient 0.533
0.603
22.86 dB / 0.726
0.512
19.96 dB / 0.597
0.484
19.07 dB / 0.553
✓ Certified Fried, JOSA 1977
8 Zernike LS + gradient 0.482
0.497
19.17 dB / 0.558
0.487
19.72 dB / 0.585
0.462
18.48 dB / 0.524
✓ Certified Noll, JOSA 1976

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 5 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 DiffusionAO + gradient 0.805 33.44 0.956
2 AO-ViT + gradient 0.767 31.2 0.933
3 AO-Transformer + gradient 0.751 30.14 0.919
4 LIFT-Net + gradient 0.729 29.07 0.901
5 WFNet + gradient 0.698 27.15 0.862
6 PnP-ADMM (WF) + gradient 0.640 24.61 0.79
7 Fried Estimator + gradient 0.603 22.86 0.726
8 Zernike LS + gradient 0.497 19.17 0.558
Spec Ranges (4 parameters)
Parameter Min Max Unit
dm_actuator_gain 0.98 1.04 -
wfs_centroid_bias -0.04 0.08 px
fried_parameter_r0 0.13 0.19 m
servo_lag -0.4 0.8 ms
View Details →
Dev 5 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 AO-ViT + gradient 0.716 28.32 0.887
2 AO-Transformer + gradient 0.679 27.3 0.865
3 DiffusionAO + gradient 0.654 25.11 0.806
4 WFNet + gradient 0.628 24.59 0.789
5 LIFT-Net + gradient 0.622 24.03 0.77
6 PnP-ADMM (WF) + gradient 0.533 21.1 0.65
7 Fried Estimator + gradient 0.512 19.96 0.597
8 Zernike LS + gradient 0.487 19.72 0.585
Spec Ranges (4 parameters)
Parameter Min Max Unit
dm_actuator_gain 0.976 1.036 -
wfs_centroid_bias -0.048 0.072 px
fried_parameter_r0 0.126 0.186 m
servo_lag -0.48 0.72 ms
Submit Reconstruction View Details →
Hidden 5 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 AO-ViT + gradient 0.645 25.09 0.805
2 AO-Transformer + gradient 0.581 22.48 0.71
3 WFNet + gradient 0.571 22.63 0.716
4 LIFT-Net + gradient 0.563 22.7 0.719
5 DiffusionAO + gradient 0.562 21.83 0.683
6 PnP-ADMM (WF) + gradient 0.489 19.52 0.575
7 Fried Estimator + gradient 0.484 19.07 0.553
8 Zernike LS + gradient 0.462 18.48 0.524
Spec Ranges (4 parameters)
Parameter Min Max Unit
dm_actuator_gain 0.986 1.046 -
wfs_centroid_bias -0.028 0.092 px
fried_parameter_r0 0.136 0.196 m
servo_lag -0.28 0.92 ms
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 → C → D

M Modulation
C Convolution
D Detector

Mismatch Parameters

Symbol Parameter Description Nominal Perturbed
d_a dm_actuator_gain DM actuator gain (-) 1.0 1.02
w_c wfs_centroid_bias WFS centroid bias (px) 0.0 0.04
f_p fried_parameter_r0 Fried parameter r0 (m) 0.15 0.17
s_l servo_lag Servo lag (ms) 0.0 0.4

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.