Entangled Photon Microscopy

Entangled Photon Microscopy

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
🥇 DiffGhost 0.872 38.8 0.950 ✓ Certified Gao et al. 2024
🥈 PhysGhost 0.836 37.1 0.936 ✓ Certified Chen et al. 2024
🥉 SwinGhost 0.803 35.6 0.920 ✓ Certified Wang et al. 2023
4 TransGhost 0.762 33.8 0.897 ✓ Certified Li et al. 2022
5 GAN-Ghost 0.693 31.0 0.852 ✓ Certified Wang et al. 2019
6 DnCNN-Ghost 0.625 28.3 0.806 ✓ Certified Lyu et al. 2017
7 SVD-Ghost 0.542 25.1 0.748 ✓ Certified Gong et al. 2010
8 CS-Ghost 0.477 22.5 0.704 ✓ Certified Katz et al. 2009
9 Coincidence-Count 0.409 19.8 0.658 ✓ Certified Pittman et al. 1995

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
🥇 SwinGhost + gradient 0.761
0.792
33.52 dB / 0.957
0.763
32.13 dB / 0.944
0.728
29.96 dB / 0.916
✓ Certified Wang et al., npj Quantum Inf. 2023
🥈 DiffGhost + gradient 0.738
0.831
36.32 dB / 0.975
0.732
29.68 dB / 0.912
0.650
25.95 dB / 0.831
✓ Certified Gao et al., NeurIPS 2024
🥉 TransGhost + gradient 0.712
0.765
31.32 dB / 0.935
0.701
27.8 dB / 0.877
0.669
26.88 dB / 0.855
✓ Certified Li et al., Opt. Express 2022
4 PhysGhost + gradient 0.700
0.810
34.46 dB / 0.964
0.694
27.14 dB / 0.862
0.596
23.96 dB / 0.767
✓ Certified Chen et al., Phys. Rev. Lett. 2024
5 GAN-Ghost + gradient 0.632
0.725
29.02 dB / 0.901
0.622
24.44 dB / 0.784
0.548
22.17 dB / 0.697
✓ Certified Wang et al., Phys. Rev. A 2019
6 DnCNN-Ghost + gradient 0.579
0.670
26.06 dB / 0.834
0.542
21.56 dB / 0.671
0.525
20.9 dB / 0.641
✓ Certified Lyu et al., Optica 2017
7 SVD-Ghost + gradient 0.457
0.583
22.11 dB / 0.695
0.448
17.99 dB / 0.500
0.341
14.26 dB / 0.321
✓ Certified Gong et al., Sci. Rep. 2010
8 Coincidence-Count + gradient 0.420
0.436
17.28 dB / 0.464
0.416
16.66 dB / 0.433
0.407
16.95 dB / 0.448
✓ Certified Pittman et al., Phys. Rev. A 1995
9 CS-Ghost + gradient 0.409
0.509
19.57 dB / 0.578
0.404
16.53 dB / 0.427
0.314
13.07 dB / 0.272
✓ Certified Katz et al., Appl. Phys. Lett. 2009

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 DiffGhost + gradient 0.831 36.32 0.975
2 PhysGhost + gradient 0.810 34.46 0.964
3 SwinGhost + gradient 0.792 33.52 0.957
4 TransGhost + gradient 0.765 31.32 0.935
5 GAN-Ghost + gradient 0.725 29.02 0.901
6 DnCNN-Ghost + gradient 0.670 26.06 0.834
7 SVD-Ghost + gradient 0.583 22.11 0.695
8 CS-Ghost + gradient 0.509 19.57 0.578
9 Coincidence-Count + gradient 0.436 17.28 0.464
Spec Ranges (4 parameters)
Parameter Min Max Unit
pair_generation_rate -2.0 4.0 -
coincidence_window -0.8 4.6 ns
accidental_coincidence_rate -4.0 8.0 -
photon_loss_(per_arm) -1.2 2.4 dB
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 SwinGhost + gradient 0.763 32.13 0.944
2 DiffGhost + gradient 0.732 29.68 0.912
3 TransGhost + gradient 0.701 27.8 0.877
4 PhysGhost + gradient 0.694 27.14 0.862
5 GAN-Ghost + gradient 0.622 24.44 0.784
6 DnCNN-Ghost + gradient 0.542 21.56 0.671
7 SVD-Ghost + gradient 0.448 17.99 0.5
8 Coincidence-Count + gradient 0.416 16.66 0.433
9 CS-Ghost + gradient 0.404 16.53 0.427
Spec Ranges (4 parameters)
Parameter Min Max Unit
pair_generation_rate -2.4 3.6 -
coincidence_window -1.16 4.24 ns
accidental_coincidence_rate -4.8 7.2 -
photon_loss_(per_arm) -1.44 2.16 dB
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 SwinGhost + gradient 0.728 29.96 0.916
2 TransGhost + gradient 0.669 26.88 0.855
3 DiffGhost + gradient 0.650 25.95 0.831
4 PhysGhost + gradient 0.596 23.96 0.767
5 GAN-Ghost + gradient 0.548 22.17 0.697
6 DnCNN-Ghost + gradient 0.525 20.9 0.641
7 Coincidence-Count + gradient 0.407 16.95 0.448
8 SVD-Ghost + gradient 0.341 14.26 0.321
9 CS-Ghost + gradient 0.314 13.07 0.272
Spec Ranges (4 parameters)
Parameter Min Max Unit
pair_generation_rate -1.4 4.6 -
coincidence_window -0.26 5.14 ns
accidental_coincidence_rate -2.8 9.2 -
photon_loss_(per_arm) -0.84 2.76 dB
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 → R → D

M Modulation
R Rotation
D Detector

Mismatch Parameters

Symbol Parameter Description Nominal Perturbed
p_g pair_generation_rate Pair generation rate (-) 0.0 2.0
c_w coincidence_window Coincidence window (ns) 1.0 2.8
a_c accidental_coincidence_rate Accidental coincidence rate (-) 0.0 4.0
p_l photon_loss_(per_arm) Photon loss (per arm) (dB) 0.0 1.2

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.