Near-field Scanning Optical Microscopy (NSOM)

Near-field Scanning Optical Microscopy (NSOM)

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
🥇 U-Net-SPM 0.775 32.94 0.952 ✓ Certified SPM U-Net variant
🥈 DiffusionSPM 0.773 32.87 0.951 ✓ Certified Zhang et al., 2024
🥉 ScoreSPM 0.750 31.8 0.940 ✓ Certified Wei et al., 2025
4 SPM-Former 0.738 31.25 0.934 ✓ Certified Chen et al., NanoLett 2024
5 E2E-BTR 0.734 31.8 0.908 ✓ Certified Kossler et al., Sci. Rep. 2022
6 DeepSPM 0.697 30.4 0.880 ✓ Certified Alldritt et al., Commun. Phys. 2020
7 TV-Deconvolution 0.609 26.32 0.841 ✓ Certified TV regularization for SPM
8 Reg-Deconv 0.582 26.8 0.770 ✓ Certified Dongmo et al., 2000
9 MLE Reconstruction 0.505 23.17 0.738 ✓ Certified Classical statistical method
10 BTR 0.452 23.2 0.630 ✓ Certified Villarrubia, JRNIST 1997

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
🥇 DiffusionSPM + gradient 0.673
0.749
30.13 dB / 0.919
0.677
26.66 dB / 0.850
0.592
23.08 dB / 0.734
✓ Certified Zhang et al., 2024
🥈 U-Net-SPM + gradient 0.672
0.753
30.65 dB / 0.926
0.666
26.26 dB / 0.839
0.596
23.21 dB / 0.739
✓ Certified SPM U-Net variant
🥉 SPM-Former + gradient 0.649
0.729
29.29 dB / 0.905
0.661
26.1 dB / 0.835
0.557
21.58 dB / 0.672
✓ Certified Chen et al., NanoLett 2024
4 E2E-BTR + gradient 0.646
0.759
30.14 dB / 0.919
0.612
23.45 dB / 0.748
0.567
22.49 dB / 0.711
✓ Certified Kossler et al., Sci. Rep. 2022
5 ScoreSPM + gradient 0.639
0.761
30.52 dB / 0.924
0.608
23.54 dB / 0.752
0.547
21.89 dB / 0.685
✓ Certified Wei et al., 2025
6 Reg-Deconv + gradient 0.575
0.625
23.83 dB / 0.762
0.565
21.88 dB / 0.685
0.535
21.58 dB / 0.672
✓ Certified Dongmo et al., 2000
7 DeepSPM + gradient 0.553
0.714
28.3 dB / 0.887
0.530
20.71 dB / 0.632
0.414
17.39 dB / 0.470
✓ Certified Alldritt et al., Commun. Phys. 2020
8 BTR + gradient 0.540
0.585
22.19 dB / 0.698
0.532
20.58 dB / 0.626
0.504
20.43 dB / 0.619
✓ Certified Villarrubia, JRNIST 1997
9 MLE Reconstruction + gradient 0.529
0.550
21.35 dB / 0.662
0.529
20.47 dB / 0.621
0.508
20.56 dB / 0.625
✓ Certified Classical statistical method
10 TV-Deconvolution + gradient 0.508
0.651
24.67 dB / 0.792
0.478
19.39 dB / 0.569
0.395
16.22 dB / 0.412
✓ Certified TV regularization for SPM

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 ScoreSPM + gradient 0.761 30.52 0.924
2 E2E-BTR + gradient 0.759 30.14 0.919
3 U-Net-SPM + gradient 0.753 30.65 0.926
4 DiffusionSPM + gradient 0.749 30.13 0.919
5 SPM-Former + gradient 0.729 29.29 0.905
6 DeepSPM + gradient 0.714 28.3 0.887
7 TV-Deconvolution + gradient 0.651 24.67 0.792
8 Reg-Deconv + gradient 0.625 23.83 0.762
9 BTR + gradient 0.585 22.19 0.698
10 MLE Reconstruction + gradient 0.550 21.35 0.662
Spec Ranges (4 parameters)
Parameter Min Max Unit
tip_sample_distance 2.0 26.0 nm
aperture_size_error -4.0 8.0 -
topographic_coupling -6.0 12.0 -
far_field_background -4.0 8.0 -
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 DiffusionSPM + gradient 0.677 26.66 0.85
2 U-Net-SPM + gradient 0.666 26.26 0.839
3 SPM-Former + gradient 0.661 26.1 0.835
4 E2E-BTR + gradient 0.612 23.45 0.748
5 ScoreSPM + gradient 0.608 23.54 0.752
6 Reg-Deconv + gradient 0.565 21.88 0.685
7 BTR + gradient 0.532 20.58 0.626
8 DeepSPM + gradient 0.530 20.71 0.632
9 MLE Reconstruction + gradient 0.529 20.47 0.621
10 TV-Deconvolution + gradient 0.478 19.39 0.569
Spec Ranges (4 parameters)
Parameter Min Max Unit
tip_sample_distance 0.4 24.4 nm
aperture_size_error -4.8 7.2 -
topographic_coupling -7.2 10.8 -
far_field_background -4.8 7.2 -
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 U-Net-SPM + gradient 0.596 23.21 0.739
2 DiffusionSPM + gradient 0.592 23.08 0.734
3 E2E-BTR + gradient 0.567 22.49 0.711
4 SPM-Former + gradient 0.557 21.58 0.672
5 ScoreSPM + gradient 0.547 21.89 0.685
6 Reg-Deconv + gradient 0.535 21.58 0.672
7 MLE Reconstruction + gradient 0.508 20.56 0.625
8 BTR + gradient 0.504 20.43 0.619
9 DeepSPM + gradient 0.414 17.39 0.47
10 TV-Deconvolution + gradient 0.395 16.22 0.412
Spec Ranges (4 parameters)
Parameter Min Max Unit
tip_sample_distance 4.4 28.4 nm
aperture_size_error -2.8 9.2 -
topographic_coupling -4.2 13.8 -
far_field_background -2.8 9.2 -
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
t_d tip_sample_distance Tip-sample distance (nm) 10.0 18.0
a_s aperture_size_error Aperture size error (-) 0.0 4.0
t_c topographic_coupling Topographic coupling (-) 0.0 6.0
f_b far_field_background Far-field background (-) 0.0 4.0

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.