Ghost 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
🥇	ScoreQuantum ScoreQuantum Wei et al., 2025 29.52 dB SSIM 0.909 Checkpoint unavailable	0.697	29.52	0.909	✓ Certified	Wei et al., 2025
🥈	Ghost-ViT Ghost-ViT Zhu et al., 2025 30.1 dB SSIM 0.885 Checkpoint unavailable	0.694	30.1	0.885	✓ Certified	Zhu et al., 2025
🥉	Quantum-ViT Quantum-ViT Quantum imaging transformer, 2024 28.78 dB SSIM 0.896 Checkpoint unavailable	0.678	28.78	0.896	✓ Certified	Quantum imaging transformer, 2024
4	Quantum-CNN Quantum-CNN Quantum imaging CNN 28.43 dB SSIM 0.890 Checkpoint unavailable	0.669	28.43	0.890	✓ Certified	Quantum imaging CNN
5	DRU-Net DRU-Net Wang et al., Sci. Rep. 2020 28.5 dB SSIM 0.840 Checkpoint unavailable	0.645	28.5	0.840	✓ Certified	Wang et al., Sci. Rep. 2020
6	DiffusionQuantum DiffusionQuantum Zhang et al., 2024 26.49 dB SSIM 0.845 Checkpoint unavailable	0.614	26.49	0.845	✓ Certified	Zhang et al., 2024
7	Bayesian CS Bayesian CS Bayesian compressed sensing 25.07 dB SSIM 0.804 Try in SpecLab →	0.570	25.07	0.804	✓ Certified	Bayesian compressed sensing
8	CS-TVAL3 CS-TVAL3 Li et al., 2014 24.8 dB SSIM 0.710 Try in SpecLab →	0.518	24.8	0.710	✓ Certified	Li et al., 2014
9	Photon Counting Photon Counting Classical baseline 22.56 dB SSIM 0.713 Try in SpecLab →	0.482	22.56	0.713	✓ Certified	Classical baseline
10	G(2)-Corr G(2)-Corr Pittman et al., PRA 1995 21.2 dB SSIM 0.550 Try in SpecLab →	0.378	21.2	0.550	✓ Certified	Pittman et al., PRA 1995

Dataset: PWM Benchmark (10 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
🥇	Ghost-ViT + gradient Ghost-ViT + gradient Zhu et al., 2025 Score 0.658 Correct & Reconstruct →	0.658	0.705 27.8 dB / 0.877	0.666 26.43 dB / 0.844	0.604 23.45 dB / 0.748	✓ Certified	Zhu et al., 2025
🥈	Quantum-ViT + gradient Quantum-ViT + gradient Quantum imaging transformer, 2024 Score 0.639 Correct & Reconstruct →	0.639	0.685 26.86 dB / 0.855	0.648 25.39 dB / 0.814	0.584 22.61 dB / 0.715	✓ Certified	Quantum imaging transformer, 2024
🥉	ScoreQuantum + gradient ScoreQuantum + gradient Wei et al., 2025 Score 0.592 Correct & Reconstruct →	0.592	0.725 28.48 dB / 0.891	0.563 22.2 dB / 0.698	0.488 19.76 dB / 0.587	✓ Certified	Wei et al., 2025
4	DRU-Net + gradient DRU-Net + gradient Wang et al., Sci. Rep. 2020 Score 0.570 Correct & Reconstruct →	0.570	0.678 26.63 dB / 0.849	0.542 21.05 dB / 0.648	0.491 19.98 dB / 0.598	✓ Certified	Wang et al., Sci. Rep. 2020
5	Bayesian CS + gradient Bayesian CS + gradient Bayesian compressed sensing Score 0.550 Correct & Reconstruct →	0.550	0.600 23.16 dB / 0.737	0.546 21.5 dB / 0.668	0.504 20.07 dB / 0.602	✓ Certified	Bayesian compressed sensing
6	Quantum-CNN + gradient Quantum-CNN + gradient Quantum imaging CNN Score 0.543 Correct & Reconstruct →	0.543	0.676 26.32 dB / 0.841	0.509 20.18 dB / 0.607	0.443 18.25 dB / 0.512	✓ Certified	Quantum imaging CNN
7	CS-TVAL3 + gradient CS-TVAL3 + gradient Li et al., 2014 Score 0.540 Correct & Reconstruct →	0.540	0.581 22.17 dB / 0.697	0.535 21.18 dB / 0.654	0.505 20.13 dB / 0.605	✓ Certified	Li et al., 2014
8	Photon Counting + gradient Photon Counting + gradient Classical baseline Score 0.514 Correct & Reconstruct →	0.514	0.560 21.12 dB / 0.651	0.522 20.64 dB / 0.629	0.460 19.07 dB / 0.553	✓ Certified	Classical baseline
9	DiffusionQuantum + gradient DiffusionQuantum + gradient Zhang et al., 2024 Score 0.447 Correct & Reconstruct →	0.447	0.628 24.12 dB / 0.773	0.387 15.67 dB / 0.386	0.326 13.91 dB / 0.306	✓ Certified	Zhang et al., 2024
10	G(2)-Corr + gradient G(2)-Corr + gradient Pittman et al., PRA 1995 Score 0.439 Correct & Reconstruct →	0.439	0.483 18.9 dB / 0.545	0.448 18.13 dB / 0.506	0.387 15.73 dB / 0.388	✓ Certified	Pittman et al., PRA 1995

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	ScoreQuantum + gradient	0.725	28.48	0.891
2	Ghost-ViT + gradient	0.705	27.8	0.877
3	Quantum-ViT + gradient	0.685	26.86	0.855
4	DRU-Net + gradient	0.678	26.63	0.849
5	Quantum-CNN + gradient	0.676	26.32	0.841
6	DiffusionQuantum + gradient	0.628	24.12	0.773
7	Bayesian CS + gradient	0.600	23.16	0.737
8	CS-TVAL3 + gradient	0.581	22.17	0.697
9	Photon Counting + gradient	0.560	21.12	0.651
10	G(2)-Corr + gradient	0.483	18.9	0.545

Spec Ranges (4 parameters)

Parameter	Min	Max	Unit
bucket_detector_efficiency	0.8	1.1	-
speckle_correlation_mismatch	-2.0	4.0	-
background_counts	-1.0	2.0	-
number_of_measurements	-8000.0	46000.0	-

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	Ghost-ViT + gradient	0.666	26.43	0.844
2	Quantum-ViT + gradient	0.648	25.39	0.814
3	ScoreQuantum + gradient	0.563	22.2	0.698
4	Bayesian CS + gradient	0.546	21.5	0.668
5	DRU-Net + gradient	0.542	21.05	0.648
6	CS-TVAL3 + gradient	0.535	21.18	0.654
7	Photon Counting + gradient	0.522	20.64	0.629
8	Quantum-CNN + gradient	0.509	20.18	0.607
9	G(2)-Corr + gradient	0.448	18.13	0.506
10	DiffusionQuantum + gradient	0.387	15.67	0.386

Spec Ranges (4 parameters)

Parameter	Min	Max	Unit
bucket_detector_efficiency	0.82	1.12	-
speckle_correlation_mismatch	-2.4	3.6	-
background_counts	-1.2	1.8	-
number_of_measurements	-11600.0	42400.0	-

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	Ghost-ViT + gradient	0.604	23.45	0.748
2	Quantum-ViT + gradient	0.584	22.61	0.715
3	CS-TVAL3 + gradient	0.505	20.13	0.605
4	Bayesian CS + gradient	0.504	20.07	0.602
5	DRU-Net + gradient	0.491	19.98	0.598
6	ScoreQuantum + gradient	0.488	19.76	0.587
7	Photon Counting + gradient	0.460	19.07	0.553
8	Quantum-CNN + gradient	0.443	18.25	0.512
9	G(2)-Corr + gradient	0.387	15.73	0.388
10	DiffusionQuantum + gradient	0.326	13.91	0.306

Spec Ranges (4 parameters)

Parameter	Min	Max	Unit
bucket_detector_efficiency	0.77	1.07	-
speckle_correlation_mismatch	-1.4	4.6	-
background_counts	-0.7	2.3	-
number_of_measurements	-2600.0	51400.0	-

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 → Σ → D

M Modulation

Σ Summation

D Detector

Mismatch Parameters

Symbol	Parameter	Description	Nominal	Perturbed
b_d	bucket_detector_efficiency	Bucket detector efficiency (-)	1.0	0.9
s_c	speckle_correlation_mismatch	Speckle correlation mismatch (-)	0.0	2.0
b_c	background_counts	Background counts (-)	0.0	1.0
n_o	number_of_measurements	Number of measurements (-)	10000.0	28000.0

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.