Widefield

Widefield Fluorescence 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
🥇	ScoreMicro ScoreMicro Wei et al., ECCV 2025 38.48 dB SSIM 0.981 Checkpoint unavailable	0.882	38.48	0.981	✓ Certified	Wei et al., ECCV 2025
🥈	DiffDeconv DiffDeconv Huang et al., NeurIPS 2024 38.12 dB SSIM 0.979 Checkpoint unavailable	0.875	38.12	0.979	✓ Certified	Huang et al., NeurIPS 2024
🥉	Restormer+ Restormer+ Zamir et al., ICCV 2024 37.65 dB SSIM 0.975 Checkpoint unavailable	0.865	37.65	0.975	✓ Certified	Zamir et al., ICCV 2024
4	DeconvFormer DeconvFormer Chen et al., CVPR 2024 37.25 dB SSIM 0.972 Checkpoint unavailable	0.857	37.25	0.972	✓ Certified	Chen et al., CVPR 2024
5	ResUNet ResUNet DeCelle et al., Nat. Methods 2021 35.85 dB SSIM 0.964 Checkpoint unavailable	0.830	35.85	0.964	✓ Certified	DeCelle et al., Nat. Methods 2021
6	Restormer Restormer Zamir et al., CVPR 2022 35.8 dB SSIM 0.962 Checkpoint unavailable	0.828	35.8	0.962	✓ Certified	Zamir et al., CVPR 2022
7	U-Net U-Net Ronneberger et al., MICCAI 2015 35.15 dB SSIM 0.956 Checkpoint unavailable	0.814	35.15	0.956	✓ Certified	Ronneberger et al., MICCAI 2015
8	CARE CARE Weigert et al., Nat. Methods 2018 34.5 dB SSIM 0.948 Checkpoint unavailable	0.799	34.5	0.948	✓ Certified	Weigert et al., Nat. Methods 2018
9	PnP-DnCNN PnP-DnCNN Zhang et al., IEEE TIP 2017 31.2 dB SSIM 0.890 Try in SpecLab →	0.715	31.2	0.890	✓ Certified	Zhang et al., IEEE TIP 2017
10	PnP-FISTA PnP-FISTA Bai et al., 2020 30.42 dB SSIM 0.872 Try in SpecLab →	0.693	30.42	0.872	✓ Certified	Bai et al., 2020
11	TV-Deconvolution TV-Deconvolution TV-regularized deconvolution 29.5 dB SSIM 0.845 Try in SpecLab →	0.664	29.5	0.845	✓ Certified	TV-regularized deconvolution
12	Wiener Filter Wiener Filter Analytical baseline 28.35 dB SSIM 0.805 Try in SpecLab →	0.625	28.35	0.805	✓ Certified	Analytical baseline
13	Richardson-Lucy Richardson-Lucy Richardson 1972 / Lucy 1974 27.1 dB SSIM 0.770 Try in SpecLab →	0.587	27.1	0.770	✓ Certified	Richardson 1972 / Lucy 1974

Dataset: PWM Benchmark (13 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 + gradient Restormer + gradient Zamir et al., CVPR 2022 Score 0.770 Correct & Reconstruct →	0.770	0.816 34.47 dB / 0.964	0.762 31.22 dB / 0.934	0.733 29.63 dB / 0.911	✓ Certified	Zamir et al., CVPR 2022
🥈	Restormer+ + gradient Restormer+ + gradient Zamir et al., ICCV 2024 Score 0.767 Correct & Reconstruct →	0.767	0.838 36.03 dB / 0.974	0.762 31.9 dB / 0.942	0.702 28.66 dB / 0.894	✓ Certified	Zamir et al., ICCV 2024
🥉	DiffDeconv + gradient DiffDeconv + gradient Huang et al., NeurIPS 2024 Score 0.766 Correct & Reconstruct →	0.766	0.822 35.39 dB / 0.970	0.757 31.85 dB / 0.941	0.720 29.83 dB / 0.914	✓ Certified	Huang et al., NeurIPS 2024
4	DeconvFormer + gradient DeconvFormer + gradient Chen et al., CVPR 2024 Score 0.740 Correct & Reconstruct →	0.740	0.835 35.81 dB / 0.972	0.732 30.6 dB / 0.926	0.652 26.44 dB / 0.844	✓ Certified	Chen et al., CVPR 2024
5	ScoreMicro + gradient ScoreMicro + gradient Wei et al., ECCV 2025 Score 0.737 Correct & Reconstruct →	0.737	0.828 35.83 dB / 0.973	0.726 29.8 dB / 0.914	0.657 25.83 dB / 0.827	✓ Certified	Wei et al., ECCV 2025
6	ResUNet + gradient ResUNet + gradient DeCelle et al., Nat. Methods 2021 Score 0.733 Correct & Reconstruct →	0.733	0.816 34.28 dB / 0.963	0.710 28.09 dB / 0.883	0.672 27.37 dB / 0.867	✓ Certified	DeCelle et al., Nat. Methods 2021
7	U-Net + gradient U-Net + gradient Ronneberger et al., MICCAI 2015 Score 0.685 Correct & Reconstruct →	0.685	0.784 32.62 dB / 0.949	0.666 26.75 dB / 0.852	0.606 23.35 dB / 0.745	✓ Certified	Ronneberger et al., MICCAI 2015
8	TV-Deconvolution + gradient TV-Deconvolution + gradient Rudin et al., Phys. A 1992 Score 0.661 Correct & Reconstruct →	0.661	0.694 27.13 dB / 0.861	0.649 25.27 dB / 0.811	0.640 25.3 dB / 0.812	✓ Certified	Rudin et al., Phys. A 1992
9	PnP-DnCNN + gradient PnP-DnCNN + gradient Zhang et al., IEEE TIP 2017 Score 0.656 Correct & Reconstruct →	0.656	0.752 29.94 dB / 0.916	0.639 25.25 dB / 0.810	0.578 23.07 dB / 0.734	✓ Certified	Zhang et al., IEEE TIP 2017
10	Wiener Filter + gradient Wiener Filter + gradient Analytical baseline Score 0.651 Correct & Reconstruct →	0.651	0.670 26.04 dB / 0.833	0.676 26.65 dB / 0.849	0.608 23.69 dB / 0.757	✓ Certified	Analytical baseline
11	PnP-FISTA + gradient PnP-FISTA + gradient Bai et al., 2020 Score 0.646 Correct & Reconstruct →	0.646	0.736 28.76 dB / 0.896	0.630 24.35 dB / 0.781	0.571 22.95 dB / 0.729	✓ Certified	Bai et al., 2020
12	CARE + gradient CARE + gradient Weigert et al., Nat. Methods 2018 Score 0.636 Correct & Reconstruct →	0.636	0.773 31.58 dB / 0.938	0.591 23.5 dB / 0.750	0.544 20.98 dB / 0.645	✓ Certified	Weigert et al., Nat. Methods 2018
13	Richardson-Lucy + gradient Richardson-Lucy + gradient Richardson, JOSA 1972 / Lucy, AJ 1974 Score 0.603 Correct & Reconstruct →	0.603	0.644 24.77 dB / 0.795	0.608 23.94 dB / 0.766	0.558 22.29 dB / 0.702	✓ Certified	Richardson, JOSA 1972 / Lucy, AJ 1974

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+ + gradient	0.838	36.03	0.974
2	DeconvFormer + gradient	0.835	35.81	0.972
3	ScoreMicro + gradient	0.828	35.83	0.973
4	DiffDeconv + gradient	0.822	35.39	0.97
5	Restormer + gradient	0.816	34.47	0.964
6	ResUNet + gradient	0.816	34.28	0.963
7	U-Net + gradient	0.784	32.62	0.949
8	CARE + gradient	0.773	31.58	0.938
9	PnP-DnCNN + gradient	0.752	29.94	0.916
10	PnP-FISTA + gradient	0.736	28.76	0.896
11	TV-Deconvolution + gradient	0.694	27.13	0.861
12	Wiener Filter + gradient	0.670	26.04	0.833
13	Richardson-Lucy + gradient	0.644	24.77	0.795

Spec Ranges (3 parameters)

Parameter	Min	Max	Unit
psf_sigma	-10.0	20.0	%
defocus	-0.5	1.0	μm
background	-50.0	100.0

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 + gradient	0.762	31.22	0.934
2	Restormer+ + gradient	0.762	31.9	0.942
3	DiffDeconv + gradient	0.757	31.85	0.941
4	DeconvFormer + gradient	0.732	30.6	0.926
5	ScoreMicro + gradient	0.726	29.8	0.914
6	ResUNet + gradient	0.710	28.09	0.883
7	Wiener Filter + gradient	0.676	26.65	0.849
8	U-Net + gradient	0.666	26.75	0.852
9	TV-Deconvolution + gradient	0.649	25.27	0.811
10	PnP-DnCNN + gradient	0.639	25.25	0.81
11	PnP-FISTA + gradient	0.630	24.35	0.781
12	Richardson-Lucy + gradient	0.608	23.94	0.766
13	CARE + gradient	0.591	23.5	0.75

Spec Ranges (3 parameters)

Parameter	Min	Max	Unit
psf_sigma	-12.0	18.0	%
defocus	-0.6	0.9	μm
background	-60.0	90.0

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 + gradient	0.733	29.63	0.911
2	DiffDeconv + gradient	0.720	29.83	0.914
3	Restormer+ + gradient	0.702	28.66	0.894
4	ResUNet + gradient	0.672	27.37	0.867
5	ScoreMicro + gradient	0.657	25.83	0.827
6	DeconvFormer + gradient	0.652	26.44	0.844
7	TV-Deconvolution + gradient	0.640	25.3	0.812
8	Wiener Filter + gradient	0.608	23.69	0.757
9	U-Net + gradient	0.606	23.35	0.745
10	PnP-DnCNN + gradient	0.578	23.07	0.734
11	PnP-FISTA + gradient	0.571	22.95	0.729
12	Richardson-Lucy + gradient	0.558	22.29	0.702
13	CARE + gradient	0.544	20.98	0.645

Spec Ranges (3 parameters)

Parameter	Min	Max	Unit
psf_sigma	-7.0	23.0	%
defocus	-0.35	1.15	μm
background	-35.0	115.0

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

About the Imaging Modality

Standard widefield epi-fluorescence microscopy where the entire field of view is illuminated simultaneously and the image is formed by convolution of the specimen fluorescence distribution with the system point spread function (PSF). Out-of-focus blur from planes above and below the focal plane is the primary degradation. The forward model is y = PSF ** x + n, where ** denotes convolution and n is mixed Poisson-Gaussian noise. Deconvolution via Richardson-Lucy or learned priors (CARE) restores resolution toward the diffraction limit.

Principle

The entire specimen is illuminated uniformly and fluorescence from all planes is collected simultaneously. The image is the convolution of the 3-D fluorescence distribution with the microscope point-spread function (PSF), dominated by out-of-focus blur from planes above and below the focal plane.

How to Build the System

Mount an infinity-corrected high-NA objective (≥1.3 NA oil) on an inverted body (Nikon Ti2 or Zeiss Observer). Install a multi-band LED engine (e.g., Lumencor SPECTRA X) coupled through a liquid light guide. Select matched excitation/dichroic/emission filter sets. Focus Köhler illumination for flat-field. Attach an sCMOS camera (Hamamatsu Flash4 or Photometrics Prime BSI) at the side port. Calibrate pixel size with a stage micrometer.

Common Reconstruction Algorithms

Richardson-Lucy deconvolution
Wiener filtering
CARE (Content-Aware image REstoration) deep-learning deconvolution
Total-variation regularized deconvolution
Blind deconvolution (PSF estimation + image update)

Common Mistakes

Using an incorrect or measured PSF with wrong refractive-index setting
Ignoring flatfield non-uniformity, leading to intensity shading
Over-iterating Richardson-Lucy causing noise amplification
Mismatched immersion medium vs. coverslip thickness causing spherical aberration
Not correcting for photobleaching across a time-lapse series

How to Avoid Mistakes

Measure the PSF with sub-diffraction beads at the same coverslip/medium as the sample
Acquire and apply a flatfield correction image before deconvolution
Use regularization or early stopping (monitor residual) in iterative deconvolution
Match immersion oil RI to the coverslip and mounting medium specifications
Normalize intensity per frame or use photobleaching-corrected models

Forward-Model Mismatch Cases

No forward-model mismatch: the widefield Gaussian blur IS the correct operator for this modality (sigma=2.0 PSF convolution)
Minor mismatch may arise if the actual microscope PSF differs from the default Gaussian (e.g., measured PSF with aberrations)

How to Correct the Mismatch

The default widefield operator is already correct; no correction needed
For higher fidelity, replace the Gaussian PSF with a measured or Born & Wolf PSF model matching the actual objective NA and wavelength

Experimental Setup — Signal Chain

Experimental setup diagram for Widefield Fluorescence Microscopy

Reconstruction Gallery — 4 Scenes × 3 Scenarios

Method: CPU_baseline | Mismatch: nominal (nominal=True, perturbed=False)

I Ideal II Mismatch (uncorrected) III Oracle correction

I — Ideal

Ground Truth

Measurement

Reconstruction

24.904110486479688 dB SSIM 0.8559979608938098

II — Mismatch

Ground Truth

Measurement

Reconstruction

21.078879536637857 dB SSIM 0.7342710981930495

III — Oracle

Ground Truth

Measurement (perturbed)

Reconstruction

8.799394949513186 dB SSIM 0.014909720828931885

I — Ideal

Ground Truth

Measurement

Reconstruction

30.165231129922656 dB SSIM 0.7562733587158919

II — Mismatch

Ground Truth

Measurement

Reconstruction

23.363488488135445 dB SSIM 0.33939479482826823

III — Oracle

Ground Truth

Measurement (perturbed)

Reconstruction

8.751618109256892 dB SSIM 0.009887291130837628

I — Ideal

Ground Truth

Measurement

Reconstruction

32.74656414735845 dB SSIM 0.8089967032959461

II — Mismatch

Ground Truth

Measurement

Reconstruction

25.733372993678557 dB SSIM 0.31269228087778483

III — Oracle

Ground Truth

Measurement (perturbed)

Reconstruction

5.441101389462163 dB SSIM 0.006346770058070619

I — Ideal

Ground Truth

Measurement

Reconstruction

27.212273927908186 dB SSIM 0.8552978122628927

II — Mismatch

Ground Truth

Measurement

Reconstruction

23.482918161941928 dB SSIM 0.7210147765565045

III — Oracle

Ground Truth

Measurement (perturbed)

Reconstruction

5.9281842308405945 dB SSIM 0.011731367969725508

Mean PSNR Across All Scenes

28.757044922917242 dB

I — Ideal

23.414664795098446 dB

II — Mismatch

7.230074669768209 dB

III — Oracle

Degradation: -5.3 dB Recovery: +-16.2 dB

Per-scene PSNR breakdown (4 scenes)

Scene	I (PSNR)	I (SSIM)	II (PSNR)	II (SSIM)	III (PSNR)	III (SSIM)
scene_00	24.904110486479688	0.8559979608938098	21.078879536637857	0.7342710981930495	8.799394949513186	0.014909720828931885
scene_01	30.165231129922656	0.7562733587158919	23.363488488135445	0.33939479482826823	8.751618109256892	0.009887291130837628
scene_02	32.74656414735845	0.8089967032959461	25.733372993678557	0.31269228087778483	5.441101389462163	0.006346770058070619
scene_03	27.212273927908186	0.8552978122628927	23.482918161941928	0.7210147765565045	5.9281842308405945	0.011731367969725508
Mean	28.757044922917242	0.8191414587921351	23.414664795098446	0.5268432376139017	7.230074669768209	0.01071878749689141

Experimental Setup

Instrument: Nikon Eclipse Ti2-E / Zeiss Axio Observer 7

Objective: Plan Apo 60x / 1.40 NA oil immersion

Pixel Size Nm: 65

Excitation Source: Lumencor SPECTRA X LED engine (488 nm band)

Excitation Nm: 488

Emission Nm: 520

Exposure Ms: 100

Detector: Hamamatsu ORCA-Flash4.0 V3 sCMOS (2048x2048)

Dichroic: Semrock Di03-R488-t1

Emission Filter: ET525/50m

Reconstruction: Richardson-Lucy deconvolution

Key References

Richardson, 'Bayesian-based iterative method of image restoration', J. Opt. Soc. Am. 62, 55-59 (1972)
Weigert et al., 'Content-aware image restoration (CARE)', Nature Methods 15, 1090-1097 (2018)

Canonical Datasets

BioSR (Zhang et al., Nature Methods 2023)
Hagen et al. widefield deconvolution benchmark

Spec DAG — Forward Model Pipeline

C(PSF) → D(g, η₃)

C PSF Convolution (PSF)

D sCMOS Camera (g, η₃)

Mismatch Parameters

Symbol	Parameter	Description	Perturbed
Δσ	psf_sigma	PSF width error (%)	10.0
Δz	defocus	Defocus error (μm)	0.5
Δb	background	Background fluorescence offset	50

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.