Magnetic Force Microscopy (MFM)

Magnetic Force Microscopy (MFM)

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
🥇 SPM-Former 0.793 33.79 0.959 ✓ Certified Chen et al., NanoLett 2024
🥈 E2E-BTR 0.734 31.8 0.908 ✓ Certified Kossler et al., Sci. Rep. 2022
🥉 U-Net-SPM 0.715 30.29 0.921 ✓ Certified SPM U-Net variant
4 DiffusionSPM 0.709 30.01 0.917 ✓ Certified Zhang et al., 2024
5 DeepSPM 0.697 30.4 0.880 ✓ Certified Alldritt et al., Commun. Phys. 2020
6 ScoreSPM 0.680 28.85 0.898 ✓ Certified Wei et al., 2025
7 TV-Deconvolution 0.640 27.38 0.867 ✓ Certified TV regularization for SPM
8 Reg-Deconv 0.582 26.8 0.770 ✓ Certified Dongmo et al., 2000
9 MLE Reconstruction 0.538 24.12 0.773 ✓ 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
🥇 SPM-Former + gradient 0.696
0.789
32.46 dB / 0.947
0.682
26.72 dB / 0.851
0.617
23.81 dB / 0.762
✓ Certified Chen et al., NanoLett 2024
🥈 U-Net-SPM + gradient 0.639
0.713
28.37 dB / 0.888
0.623
24.48 dB / 0.785
0.581
22.8 dB / 0.723
✓ Certified SPM U-Net variant
🥉 E2E-BTR + gradient 0.614
0.729
28.82 dB / 0.897
0.592
23.39 dB / 0.746
0.522
20.31 dB / 0.613
✓ Certified Kossler et al., Sci. Rep. 2022
4 DeepSPM + gradient 0.610
0.734
28.77 dB / 0.896
0.592
23.15 dB / 0.737
0.503
19.64 dB / 0.581
✓ Certified Alldritt et al., Commun. Phys. 2020
5 DiffusionSPM + gradient 0.573
0.702
27.45 dB / 0.869
0.545
21.09 dB / 0.650
0.473
18.7 dB / 0.535
✓ Certified Zhang et al., 2024
6 TV-Deconvolution + gradient 0.555
0.682
26.13 dB / 0.836
0.518
19.96 dB / 0.597
0.464
18.86 dB / 0.543
✓ Certified TV regularization for SPM
7 MLE Reconstruction + gradient 0.553
0.570
21.93 dB / 0.687
0.569
22.27 dB / 0.701
0.519
20.9 dB / 0.641
✓ Certified Classical statistical method
8 ScoreSPM + gradient 0.538
0.680
26.43 dB / 0.844
0.485
19.09 dB / 0.554
0.448
18.35 dB / 0.517
✓ Certified Wei et al., 2025
9 Reg-Deconv + gradient 0.530
0.640
24.78 dB / 0.795
0.524
20.57 dB / 0.626
0.426
17.19 dB / 0.460
✓ Certified Dongmo et al., 2000
10 BTR + gradient 0.521
0.532
20.44 dB / 0.620
0.538
20.71 dB / 0.632
0.494
20.23 dB / 0.610
✓ Certified Villarrubia, JRNIST 1997

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 SPM-Former + gradient 0.789 32.46 0.947
2 DeepSPM + gradient 0.734 28.77 0.896
3 E2E-BTR + gradient 0.729 28.82 0.897
4 U-Net-SPM + gradient 0.713 28.37 0.888
5 DiffusionSPM + gradient 0.702 27.45 0.869
6 TV-Deconvolution + gradient 0.682 26.13 0.836
7 ScoreSPM + gradient 0.680 26.43 0.844
8 Reg-Deconv + gradient 0.640 24.78 0.795
9 MLE Reconstruction + gradient 0.570 21.93 0.687
10 BTR + gradient 0.532 20.44 0.62
Spec Ranges (3 parameters)
Parameter Min Max Unit
lift_height 20.0 110.0 nm
tip_magnetization_model -0.15 0.15 -
electrostatic_coupling -2.0 4.0 -
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 SPM-Former + gradient 0.682 26.72 0.851
2 U-Net-SPM + gradient 0.623 24.48 0.785
3 E2E-BTR + gradient 0.592 23.39 0.746
4 DeepSPM + gradient 0.592 23.15 0.737
5 MLE Reconstruction + gradient 0.569 22.27 0.701
6 DiffusionSPM + gradient 0.545 21.09 0.65
7 BTR + gradient 0.538 20.71 0.632
8 Reg-Deconv + gradient 0.524 20.57 0.626
9 TV-Deconvolution + gradient 0.518 19.96 0.597
10 ScoreSPM + gradient 0.485 19.09 0.554
Spec Ranges (3 parameters)
Parameter Min Max Unit
lift_height 14.0 104.0 nm
tip_magnetization_model -0.15 0.15 -
electrostatic_coupling -2.4 3.6 -
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 SPM-Former + gradient 0.617 23.81 0.762
2 U-Net-SPM + gradient 0.581 22.8 0.723
3 E2E-BTR + gradient 0.522 20.31 0.613
4 MLE Reconstruction + gradient 0.519 20.9 0.641
5 DeepSPM + gradient 0.503 19.64 0.581
6 BTR + gradient 0.494 20.23 0.61
7 DiffusionSPM + gradient 0.473 18.7 0.535
8 TV-Deconvolution + gradient 0.464 18.86 0.543
9 ScoreSPM + gradient 0.448 18.35 0.517
10 Reg-Deconv + gradient 0.426 17.19 0.46
Spec Ranges (3 parameters)
Parameter Min Max Unit
lift_height 29.0 119.0 nm
tip_magnetization_model -0.15 0.15 -
electrostatic_coupling -1.4 4.6 -
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

S → M → D

S Sampling
M Modulation
D Detector

Mismatch Parameters

Symbol Parameter Description Nominal Perturbed
l_h lift_height Lift height (nm) 50.0 80.0
t_m tip_magnetization_model Tip magnetization model (-) 0.0 0.0
e_c electrostatic_coupling Electrostatic coupling (-) 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.