Streak Camera 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
🥇	UltraFormer UltraFormer Ultrafast transformer, 2024 34.63 dB SSIM 0.965 Checkpoint unavailable	0.810	34.63	0.965	✓ Certified	Ultrafast transformer, 2024
🥈	Temporal-U-Net Temporal-U-Net 3D/Temporal U-Net variant 32.94 dB SSIM 0.952 Checkpoint unavailable	0.775	32.94	0.952	✓ Certified	3D/Temporal U-Net variant
🥉	AL-DL AL-DL Yao et al., Photon. Res. 2021 33.4 dB SSIM 0.930 Checkpoint unavailable	0.772	33.4	0.930	✓ Certified	Yao et al., Photon. Res. 2021
4	CUP-Net CUP-Net Parker et al., 2021 31.9 dB SSIM 0.900 Checkpoint unavailable	0.732	31.9	0.900	✓ Certified	Parker et al., 2021
5	DiffusionUltrafast DiffusionUltrafast Zhang et al., 2024 30.14 dB SSIM 0.919 Checkpoint unavailable	0.712	30.14	0.919	✓ Certified	Zhang et al., 2024
6	Unfolded-CUP Unfolded-CUP CUP algorithm unfolding 29.78 dB SSIM 0.913 Checkpoint unavailable	0.703	29.78	0.913	✓ Certified	CUP algorithm unfolding
7	ScoreUltrafast ScoreUltrafast Wei et al., 2025 29.43 dB SSIM 0.908 Checkpoint unavailable	0.695	29.43	0.908	✓ Certified	Wei et al., 2025
8	PnP-FFDNet PnP-FFDNet Yuan et al., 2020 28.3 dB SSIM 0.820 Try in SpecLab →	0.632	28.3	0.820	✓ Certified	Yuan et al., 2020
9	Temporal Filtering Temporal Filtering Analytical baseline 26.94 dB SSIM 0.857 Try in SpecLab →	0.627	26.94	0.857	✓ Certified	Analytical baseline
10	PnP-ADMM PnP-ADMM ADMM + denoiser prior 26.91 dB SSIM 0.856 Try in SpecLab →	0.627	26.91	0.856	✓ Certified	ADMM + denoiser prior
11	TwIST TwIST Bioucas-Dias & Figueiredo, IEEE TIP 2007 24.6 dB SSIM 0.680 Try in SpecLab →	0.500	24.6	0.680	✓ Certified	Bioucas-Dias & Figueiredo, IEEE TIP 2007

Dataset: PWM Benchmark (11 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
🥇	UltraFormer + gradient UltraFormer + gradient Ultrafast transformer, 2024 Score 0.746 Correct & Reconstruct →	0.746	0.781 32.71 dB / 0.950	0.754 31.26 dB / 0.934	0.702 27.96 dB / 0.880	✓ Certified	Ultrafast transformer, 2024
🥈	AL-DL + gradient AL-DL + gradient Yao et al., Photon. Res. 2021 Score 0.668 Correct & Reconstruct →	0.668	0.763 31.59 dB / 0.938	0.639 24.84 dB / 0.797	0.601 24.16 dB / 0.774	✓ Certified	Yao et al., Photon. Res. 2021
🥉	Temporal-U-Net + gradient Temporal-U-Net + gradient 3D/Temporal U-Net variant Score 0.655 Correct & Reconstruct →	0.655	0.754 30.56 dB / 0.925	0.636 25.23 dB / 0.809	0.576 22.3 dB / 0.703	✓ Certified	3D/Temporal U-Net variant
4	Temporal Filtering + gradient Temporal Filtering + gradient Analytical baseline Score 0.632 Correct & Reconstruct →	0.632	0.667 25.36 dB / 0.813	0.644 25.21 dB / 0.809	0.585 23.5 dB / 0.750	✓ Certified	Analytical baseline
5	Unfolded-CUP + gradient Unfolded-CUP + gradient CUP algorithm unfolding Score 0.628 Correct & Reconstruct →	0.628	0.727 28.47 dB / 0.890	0.584 23.05 dB / 0.733	0.573 22.47 dB / 0.710	✓ Certified	CUP algorithm unfolding
6	CUP-Net + gradient CUP-Net + gradient Parker et al., 2021 Score 0.598 Correct & Reconstruct →	0.598	0.739 29.77 dB / 0.913	0.570 21.85 dB / 0.684	0.485 19.63 dB / 0.581	✓ Certified	Parker et al., 2021
7	PnP-FFDNet + gradient PnP-FFDNet + gradient Yuan et al., 2020 Score 0.582 Correct & Reconstruct →	0.582	0.670 26.05 dB / 0.833	0.584 22.84 dB / 0.725	0.491 19.86 dB / 0.592	✓ Certified	Yuan et al., 2020
8	PnP-ADMM + gradient PnP-ADMM + gradient ADMM + denoiser prior Score 0.580 Correct & Reconstruct →	0.580	0.637 24.48 dB / 0.785	0.584 22.43 dB / 0.708	0.519 20.38 dB / 0.617	✓ Certified	ADMM + denoiser prior
9	ScoreUltrafast + gradient ScoreUltrafast + gradient Wei et al., 2025 Score 0.570 Correct & Reconstruct →	0.570	0.688 26.75 dB / 0.852	0.548 21.05 dB / 0.648	0.475 19.46 dB / 0.572	✓ Certified	Wei et al., 2025
10	TwIST + gradient TwIST + gradient Bioucas-Dias & Figueiredo, IEEE TIP 2007 Score 0.563 Correct & Reconstruct →	0.563	0.614 23.21 dB / 0.739	0.534 20.77 dB / 0.635	0.540 21.42 dB / 0.665	✓ Certified	Bioucas-Dias & Figueiredo, IEEE TIP 2007
11	DiffusionUltrafast + gradient DiffusionUltrafast + gradient Zhang et al., 2024 Score 0.550 Correct & Reconstruct →	0.550	0.709 28.15 dB / 0.884	0.495 19.37 dB / 0.568	0.446 18.08 dB / 0.504	✓ Certified	Zhang et al., 2024

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	UltraFormer + gradient	0.781	32.71	0.95
2	AL-DL + gradient	0.763	31.59	0.938
3	Temporal-U-Net + gradient	0.754	30.56	0.925
4	CUP-Net + gradient	0.739	29.77	0.913
5	Unfolded-CUP + gradient	0.727	28.47	0.89
6	DiffusionUltrafast + gradient	0.709	28.15	0.884
7	ScoreUltrafast + gradient	0.688	26.75	0.852
8	PnP-FFDNet + gradient	0.670	26.05	0.833
9	Temporal Filtering + gradient	0.667	25.36	0.813
10	PnP-ADMM + gradient	0.637	24.48	0.785
11	TwIST + gradient	0.614	23.21	0.739

Spec Ranges (4 parameters)

Parameter	Min	Max	Unit
sweep_nonlinearity	-1.0	2.0	-
temporal_resolution	0.2	2.6	ps
dynamic_range_saturation	-2.0	4.0	-
trigger_jitter	-2.0	4.0	ps

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	UltraFormer + gradient	0.754	31.26	0.934
2	Temporal Filtering + gradient	0.644	25.21	0.809
3	AL-DL + gradient	0.639	24.84	0.797
4	Temporal-U-Net + gradient	0.636	25.23	0.809
5	Unfolded-CUP + gradient	0.584	23.05	0.733
6	PnP-FFDNet + gradient	0.584	22.84	0.725
7	PnP-ADMM + gradient	0.584	22.43	0.708
8	CUP-Net + gradient	0.570	21.85	0.684
9	ScoreUltrafast + gradient	0.548	21.05	0.648
10	TwIST + gradient	0.534	20.77	0.635
11	DiffusionUltrafast + gradient	0.495	19.37	0.568

Spec Ranges (4 parameters)

Parameter	Min	Max	Unit
sweep_nonlinearity	-1.2	1.8	-
temporal_resolution	0.04	2.44	ps
dynamic_range_saturation	-2.4	3.6	-
trigger_jitter	-2.4	3.6	ps

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	UltraFormer + gradient	0.702	27.96	0.88
2	AL-DL + gradient	0.601	24.16	0.774
3	Temporal Filtering + gradient	0.585	23.5	0.75
4	Temporal-U-Net + gradient	0.576	22.3	0.703
5	Unfolded-CUP + gradient	0.573	22.47	0.71
6	TwIST + gradient	0.540	21.42	0.665
7	PnP-ADMM + gradient	0.519	20.38	0.617
8	PnP-FFDNet + gradient	0.491	19.86	0.592
9	CUP-Net + gradient	0.485	19.63	0.581
10	ScoreUltrafast + gradient	0.475	19.46	0.572
11	DiffusionUltrafast + gradient	0.446	18.08	0.504

Spec Ranges (4 parameters)

Parameter	Min	Max	Unit
sweep_nonlinearity	-0.7	2.3	-
temporal_resolution	0.44	2.84	ps
dynamic_range_saturation	-1.4	4.6	-
trigger_jitter	-1.4	4.6	ps

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
s_n	sweep_nonlinearity	Sweep nonlinearity (-)	0.0	1.0
t_r	temporal_resolution	Temporal resolution (ps)	1.0	1.8
d_r	dynamic_range_saturation	Dynamic range saturation (-)	0.0	2.0
t_j	trigger_jitter	Trigger jitter (ps)	0.0	2.0

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.