Cathodoluminescence (CL) 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
🥇	DiffusionEM DiffusionEM Gao et al. 2024 39.8 dB SSIM 0.962 Checkpoint unavailable	0.894	39.8	0.962	✓ Certified	Gao et al. 2024
🥈	Restormer-CL Restormer-CL Zamir et al. 2022 38.4 dB SSIM 0.950 Checkpoint unavailable	0.865	38.4	0.950	✓ Certified	Zamir et al. 2022
🥉	SwinIR-CL SwinIR-CL Liang et al. 2021 37.1 dB SSIM 0.938 Checkpoint unavailable	0.837	37.1	0.938	✓ Certified	Liang et al. 2021
4	PINN-CL PINN-CL Raissi et al. 2019 36.8 dB SSIM 0.934 Checkpoint unavailable	0.830	36.8	0.934	✓ Certified	Raissi et al. 2019
5	CARE-CL CARE-CL Weigert et al. 2018 35.5 dB SSIM 0.921 Checkpoint unavailable	0.802	35.5	0.921	✓ Certified	Weigert et al. 2018
6	U-Net-CL U-Net-CL Ronneberger et al. 2015 34.2 dB SSIM 0.908 Checkpoint unavailable	0.774	34.2	0.908	✓ Certified	Ronneberger et al. 2015
7	DnCNN-CL DnCNN-CL Zhang et al. 2017 31.8 dB SSIM 0.875 Checkpoint unavailable	0.718	31.8	0.875	✓ Certified	Zhang et al. 2017
8	Richardson-Lucy Richardson-Lucy Richardson 1972 27.5 dB SSIM 0.812 Try in SpecLab →	0.614	27.5	0.812	✓ Certified	Richardson 1972
9	Wiener-CL Wiener-CL Castleman 1996 25.2 dB SSIM 0.771 Try in SpecLab →	0.555	25.2	0.771	✓ Certified	Castleman 1996

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
🥇	Restormer-CL + gradient Restormer-CL + gradient Zamir et al., CVPR 2022 (CL adapted) Score 0.802 Correct & Reconstruct →	0.802	0.847 36.66 dB / 0.977	0.797 34.87 dB / 0.967	0.762 31.76 dB / 0.940	✓ Certified	Zamir et al., CVPR 2022 (CL adapted)
🥈	PINN-CL + gradient PINN-CL + gradient Raissi et al., J. Comput. Phys. 2019 (CL) Score 0.751 Correct & Reconstruct →	0.751	0.828 35.09 dB / 0.968	0.745 30.62 dB / 0.926	0.680 27.63 dB / 0.873	✓ Certified	Raissi et al., J. Comput. Phys. 2019 (CL)
🥉	SwinIR-CL + gradient SwinIR-CL + gradient Liang et al., ICCV 2021 (CL adapted) Score 0.735 Correct & Reconstruct →	0.735	0.810 34.79 dB / 0.966	0.729 29.4 dB / 0.907	0.667 25.93 dB / 0.830	✓ Certified	Liang et al., ICCV 2021 (CL adapted)
4	DiffusionEM + gradient DiffusionEM + gradient Gao et al., Nat. Methods 2024 (EM adapted) Score 0.735 Correct & Reconstruct →	0.735	0.843 37.51 dB / 0.980	0.713 29.16 dB / 0.903	0.650 25.5 dB / 0.818	✓ Certified	Gao et al., Nat. Methods 2024 (EM adapted)
5	CARE-CL + gradient CARE-CL + gradient Weigert et al., Nat. Methods 2018 (CL adapted) Score 0.724 Correct & Reconstruct →	0.724	0.792 33.34 dB / 0.956	0.714 28.76 dB / 0.896	0.666 26.62 dB / 0.849	✓ Certified	Weigert et al., Nat. Methods 2018 (CL adapted)
6	U-Net-CL + gradient U-Net-CL + gradient Ronneberger et al., MICCAI 2015 (CL adapted) Score 0.698 Correct & Reconstruct →	0.698	0.796 32.94 dB / 0.952	0.681 27.2 dB / 0.863	0.617 24.75 dB / 0.794	✓ Certified	Ronneberger et al., MICCAI 2015 (CL adapted)
7	Richardson-Lucy + gradient Richardson-Lucy + gradient Richardson, J. Opt. Soc. Am. 1972 Score 0.654 Correct & Reconstruct →	0.654	0.680 25.95 dB / 0.831	0.642 24.77 dB / 0.795	0.640 25.3 dB / 0.812	✓ Certified	Richardson, J. Opt. Soc. Am. 1972
8	DnCNN-CL + gradient DnCNN-CL + gradient Zhang et al., IEEE TIP 2017 (CL adapted) Score 0.648 Correct & Reconstruct →	0.648	0.731 28.92 dB / 0.899	0.635 24.41 dB / 0.783	0.579 22.56 dB / 0.713	✓ Certified	Zhang et al., IEEE TIP 2017 (CL adapted)
9	Wiener-CL + gradient Wiener-CL + gradient Castleman, Digital Image Processing, 1996 Score 0.571 Correct & Reconstruct →	0.571	0.593 22.73 dB / 0.720	0.594 23.04 dB / 0.733	0.526 20.5 dB / 0.622	✓ Certified	Castleman, Digital Image Processing, 1996

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	Restormer-CL + gradient	0.847	36.66	0.977
2	DiffusionEM + gradient	0.843	37.51	0.98
3	PINN-CL + gradient	0.828	35.09	0.968
4	SwinIR-CL + gradient	0.810	34.79	0.966
5	U-Net-CL + gradient	0.796	32.94	0.952
6	CARE-CL + gradient	0.792	33.34	0.956
7	DnCNN-CL + gradient	0.731	28.92	0.899
8	Richardson-Lucy + gradient	0.680	25.95	0.831
9	Wiener-CL + gradient	0.593	22.73	0.72

Spec Ranges (4 parameters)

Parameter	Min	Max	Unit
beam_current_drift	-1.0	2.0	-
collection_efficiency_variation	-4.0	8.0	spatial
spectral_calibration_error	-0.4	0.8	nm
carbon_contamination	-2.0	4.0	signalloss

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	Restormer-CL + gradient	0.797	34.87	0.967
2	PINN-CL + gradient	0.745	30.62	0.926
3	SwinIR-CL + gradient	0.729	29.4	0.907
4	CARE-CL + gradient	0.714	28.76	0.896
5	DiffusionEM + gradient	0.713	29.16	0.903
6	U-Net-CL + gradient	0.681	27.2	0.863
7	Richardson-Lucy + gradient	0.642	24.77	0.795
8	DnCNN-CL + gradient	0.635	24.41	0.783
9	Wiener-CL + gradient	0.594	23.04	0.733

Spec Ranges (4 parameters)

Parameter	Min	Max	Unit
beam_current_drift	-1.2	1.8	-
collection_efficiency_variation	-4.8	7.2	spatial
spectral_calibration_error	-0.48	0.72	nm
carbon_contamination	-2.4	3.6	signalloss

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	Restormer-CL + gradient	0.762	31.76	0.94
2	PINN-CL + gradient	0.680	27.63	0.873
3	SwinIR-CL + gradient	0.667	25.93	0.83
4	CARE-CL + gradient	0.666	26.62	0.849
5	DiffusionEM + gradient	0.650	25.5	0.818
6	Richardson-Lucy + gradient	0.640	25.3	0.812
7	U-Net-CL + gradient	0.617	24.75	0.794
8	DnCNN-CL + gradient	0.579	22.56	0.713
9	Wiener-CL + gradient	0.526	20.5	0.622

Spec Ranges (4 parameters)

Parameter	Min	Max	Unit
beam_current_drift	-0.7	2.3	-
collection_efficiency_variation	-2.8	9.2	spatial
spectral_calibration_error	-0.28	0.92	nm
carbon_contamination	-1.4	4.6	signalloss

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	Perturbed
b_c	beam_current_drift	Beam current drift (-)	1.0
c_e	collection_efficiency_variation	Collection efficiency variation (spatial)	4.0
s_c	spectral_calibration_error	Spectral calibration error (nm)	0.4
c_c	carbon_contamination	Carbon contamination (signal loss)	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.