STEM-EDX Elemental Mapping

STEM-EDX Elemental Mapping

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
🥇 DiffEDX 0.884 39.4 0.955 ✓ Certified Gao et al. 2024
🥈 PhysEDX 0.853 37.9 0.943 ✓ Certified Chen et al. 2024
🥉 SwinEDX 0.830 36.8 0.933 ✓ Certified Wang et al. 2023
4 TransEDX 0.795 35.2 0.916 ✓ Certified Li et al. 2022
5 N2V-EDX 0.736 32.8 0.878 ✓ Certified Krull et al. 2019
6 DnCNN-EDX 0.676 30.3 0.843 ✓ Certified Kovarik et al. 2016
7 NMF-EDX 0.604 27.5 0.792 ✓ Certified Nicoletti et al. 2013
8 TV-EDX 0.540 24.9 0.751 ✓ Certified Saghi et al. 2011
9 MLS-EDX 0.476 22.3 0.708 ✓ Certified Statham 1995

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
🥇 SwinEDX + gradient 0.778
0.830
35.56 dB / 0.971
0.782
33.09 dB / 0.953
0.723
30.17 dB / 0.919
✓ Certified Wang et al., npj Comput. Mater. 2023
🥈 PhysEDX + gradient 0.772
0.842
36.7 dB / 0.977
0.758
30.85 dB / 0.929
0.716
29.83 dB / 0.914
✓ Certified Chen et al., Microsc. Microanal. 2024
🥉 TransEDX + gradient 0.753
0.809
33.93 dB / 0.960
0.742
30.58 dB / 0.925
0.709
28.74 dB / 0.895
✓ Certified Li et al., Ultramicroscopy 2022
4 DiffEDX + gradient 0.751
0.838
36.92 dB / 0.978
0.749
30.79 dB / 0.928
0.665
26.09 dB / 0.835
✓ Certified Gao et al., NeurIPS 2024
5 DnCNN-EDX + gradient 0.639
0.710
28.1 dB / 0.883
0.637
24.84 dB / 0.797
0.569
22.57 dB / 0.714
✓ Certified Kovarik et al., npj Comput. Mater. 2016
6 N2V-EDX + gradient 0.632
0.775
31.45 dB / 0.936
0.621
24.33 dB / 0.780
0.499
19.63 dB / 0.581
✓ Certified Krull et al., NeurIPS 2019
7 NMF-EDX + gradient 0.512
0.645
24.76 dB / 0.794
0.475
18.65 dB / 0.532
0.415
17.18 dB / 0.459
✓ Certified Nicoletti et al., Nature 2013
8 MLS-EDX + gradient 0.448
0.508
19.53 dB / 0.576
0.446
18.32 dB / 0.516
0.389
16.08 dB / 0.405
✓ Certified Statham, J. Anal. At. Spectrom. 1995
9 TV-EDX + gradient 0.443
0.592
22.82 dB / 0.724
0.420
16.99 dB / 0.450
0.317
13.98 dB / 0.309
✓ Certified Saghi et al., Ultramicroscopy 2011

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 PhysEDX + gradient 0.842 36.7 0.977
2 DiffEDX + gradient 0.838 36.92 0.978
3 SwinEDX + gradient 0.830 35.56 0.971
4 TransEDX + gradient 0.809 33.93 0.96
5 N2V-EDX + gradient 0.775 31.45 0.936
6 DnCNN-EDX + gradient 0.710 28.1 0.883
7 NMF-EDX + gradient 0.645 24.76 0.794
8 TV-EDX + gradient 0.592 22.82 0.724
9 MLS-EDX + gradient 0.508 19.53 0.576
Spec Ranges (4 parameters)
Parameter Min Max Unit
absorption_correction_error -3.0 6.0 -
detector_solid_angle -0.15 0.15 sr
peak_overlap_(spectral) -0.6 1.2 -
bremsstrahlung_background -0.15 0.15 -
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 SwinEDX + gradient 0.782 33.09 0.953
2 PhysEDX + gradient 0.758 30.85 0.929
3 DiffEDX + gradient 0.749 30.79 0.928
4 TransEDX + gradient 0.742 30.58 0.925
5 DnCNN-EDX + gradient 0.637 24.84 0.797
6 N2V-EDX + gradient 0.621 24.33 0.78
7 NMF-EDX + gradient 0.475 18.65 0.532
8 MLS-EDX + gradient 0.446 18.32 0.516
9 TV-EDX + gradient 0.420 16.99 0.45
Spec Ranges (4 parameters)
Parameter Min Max Unit
absorption_correction_error -3.6 5.4 -
detector_solid_angle -0.15 0.15 sr
peak_overlap_(spectral) -0.72 1.08 -
bremsstrahlung_background -0.15 0.15 -
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 SwinEDX + gradient 0.723 30.17 0.919
2 PhysEDX + gradient 0.716 29.83 0.914
3 TransEDX + gradient 0.709 28.74 0.895
4 DiffEDX + gradient 0.665 26.09 0.835
5 DnCNN-EDX + gradient 0.569 22.57 0.714
6 N2V-EDX + gradient 0.499 19.63 0.581
7 NMF-EDX + gradient 0.415 17.18 0.459
8 MLS-EDX + gradient 0.389 16.08 0.405
9 TV-EDX + gradient 0.317 13.98 0.309
Spec Ranges (4 parameters)
Parameter Min Max Unit
absorption_correction_error -2.1 6.9 -
detector_solid_angle -0.15 0.15 sr
peak_overlap_(spectral) -0.42 1.38 -
bremsstrahlung_background -0.15 0.15 -
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 Nominal Perturbed
a_c absorption_correction_error Absorption correction error (-) 0.0 3.0
d_s detector_solid_angle Detector solid angle (sr) 0.0 0.0
p_o peak_overlap_(spectral) Peak overlap (spectral) (-) 0.0 0.6
b_b bremsstrahlung_background Bremsstrahlung background (-) 0.0 0.0

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.