Coded Exposure / Flutter Shutter

Coded Exposure / Flutter Shutter

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
🥇 DiffusionDeblur 0.894 39.8 0.961 ✓ Certified Whang 2022
🥈 Restormer-Deblur 0.869 38.6 0.951 ✓ Certified Zamir 2022
🥉 MPRNet 0.844 37.4 0.941 ✓ Certified Zamir 2021
4 DMPHN 0.816 36.1 0.928 ✓ Certified Zhang 2019
5 DeblurGAN 0.787 34.8 0.914 ✓ Certified Kupyn 2018
6 DnCNN-Deblur 0.758 33.5 0.899 ✓ Certified Zhang 2017
7 BM3D-Deblur 0.716 31.8 0.871 ✓ Certified Danielyan 2012
8 TV-Deconv 0.652 29.2 0.831 ✓ Certified Rudin 1992
9 Wiener-Deconv 0.587 26.5 0.791 ✓ Certified Wiener 1942

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
🥇 DiffusionDeblur + gradient 0.756
0.844
37.35 dB / 0.980
0.744
30.56 dB / 0.925
0.679
27.61 dB / 0.872
✓ Certified Whang et al., CVPR 2022
🥈 Restormer-Deblur + gradient 0.755
0.829
36.17 dB / 0.974
0.764
32.07 dB / 0.943
0.673
26.69 dB / 0.850
✓ Certified Zamir et al., CVPR 2022
🥉 MPRNet + gradient 0.746
0.816
35.03 dB / 0.968
0.722
28.86 dB / 0.898
0.700
28.31 dB / 0.887
✓ Certified Zamir et al., CVPR 2021
4 DMPHN + gradient 0.702
0.798
33.49 dB / 0.957
0.675
26.11 dB / 0.835
0.634
25.04 dB / 0.803
✓ Certified Zhang et al., CVPR 2019
5 BM3D-Deblur + gradient 0.698
0.732
28.96 dB / 0.900
0.706
28.56 dB / 0.892
0.657
25.93 dB / 0.830
✓ Certified Danielyan et al., IEEE TIP 2012
6 DeblurGAN + gradient 0.669
0.783
32.86 dB / 0.951
0.638
24.52 dB / 0.787
0.586
23.06 dB / 0.733
✓ Certified Kupyn et al., CVPR 2018
7 DnCNN-Deblur + gradient 0.646
0.760
30.71 dB / 0.927
0.623
23.89 dB / 0.765
0.554
21.87 dB / 0.684
✓ Certified Zhang et al., IEEE TIP 2017
8 Wiener-Deconv + gradient 0.583
0.626
24.02 dB / 0.769
0.596
23.43 dB / 0.748
0.528
21.35 dB / 0.662
✓ Certified Wiener, MIT Tech. Rep. 1942
9 TV-Deconv + gradient 0.546
0.686
26.64 dB / 0.849
0.527
20.46 dB / 0.621
0.425
17.31 dB / 0.466
✓ Certified Rudin et al., Physica D 1992

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 DiffusionDeblur + gradient 0.844 37.35 0.98
2 Restormer-Deblur + gradient 0.829 36.17 0.974
3 MPRNet + gradient 0.816 35.03 0.968
4 DMPHN + gradient 0.798 33.49 0.957
5 DeblurGAN + gradient 0.783 32.86 0.951
6 DnCNN-Deblur + gradient 0.760 30.71 0.927
7 BM3D-Deblur + gradient 0.732 28.96 0.9
8 TV-Deconv + gradient 0.686 26.64 0.849
9 Wiener-Deconv + gradient 0.626 24.02 0.769
Spec Ranges (3 parameters)
Parameter Min Max Unit
shutter_code_timing_error -1.0 2.0 -
motion_blur_psf_mismatch -4.0 8.0 velocityerror
sensor_readout_noise 3.0 9.0 e-
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-Deblur + gradient 0.764 32.07 0.943
2 DiffusionDeblur + gradient 0.744 30.56 0.925
3 MPRNet + gradient 0.722 28.86 0.898
4 BM3D-Deblur + gradient 0.706 28.56 0.892
5 DMPHN + gradient 0.675 26.11 0.835
6 DeblurGAN + gradient 0.638 24.52 0.787
7 DnCNN-Deblur + gradient 0.623 23.89 0.765
8 Wiener-Deconv + gradient 0.596 23.43 0.748
9 TV-Deconv + gradient 0.527 20.46 0.621
Spec Ranges (3 parameters)
Parameter Min Max Unit
shutter_code_timing_error -1.2 1.8 -
motion_blur_psf_mismatch -4.8 7.2 velocityerror
sensor_readout_noise 2.6 8.6 e-
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 MPRNet + gradient 0.700 28.31 0.887
2 DiffusionDeblur + gradient 0.679 27.61 0.872
3 Restormer-Deblur + gradient 0.673 26.69 0.85
4 BM3D-Deblur + gradient 0.657 25.93 0.83
5 DMPHN + gradient 0.634 25.04 0.803
6 DeblurGAN + gradient 0.586 23.06 0.733
7 DnCNN-Deblur + gradient 0.554 21.87 0.684
8 Wiener-Deconv + gradient 0.528 21.35 0.662
9 TV-Deconv + gradient 0.425 17.31 0.466
Spec Ranges (3 parameters)
Parameter Min Max Unit
shutter_code_timing_error -0.7 2.3 -
motion_blur_psf_mismatch -2.8 9.2 velocityerror
sensor_readout_noise 3.6 9.6 e-
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 → C → D

M Modulation
C Convolution
D Detector

Mismatch Parameters

Symbol Parameter Description Nominal Perturbed
s_c shutter_code_timing_error Shutter code timing error (-) 0.0 1.0
m_b motion_blur_psf_mismatch Motion blur PSF mismatch (velocity error) 0.0 4.0
s_r sensor_readout_noise Sensor readout noise (e-) 5.0 7.0

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.