Cryo-Electron Tomography (Cryo-ET)

Cryo-Electron Tomography (Cryo-ET)

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
🥇 DiffusionET 0.854 37.9 0.944 ✓ Certified Zhang 2024
🥈 ETFormer 0.804 35.6 0.921 ✓ Certified Chen 2024
🥉 DeePiCt 0.775 34.2 0.909 ✓ Certified Moebel 2021
4 CryoSeg 0.751 33.1 0.898 ✓ Certified Lamm 2022
5 DeepDeWedge 0.716 31.7 0.876 ✓ Certified Wiedemann 2024
6 IsoNet 0.661 29.4 0.842 ✓ Certified Liu 2021
7 IMOD 0.557 25.2 0.774 ✓ Certified Kremer 1996
8 SART-ET 0.517 23.8 0.741 ✓ Certified Andersen 1984
9 WBP 0.433 20.5 0.682 ✓ Certified Crowther 1970

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
🥇 DiffusionET + gradient 0.719
0.842
36.21 dB / 0.974
0.698
28.41 dB / 0.889
0.618
24.1 dB / 0.772
✓ Certified Zhang et al., arXiv 2024
🥈 ETFormer + gradient 0.712
0.813
34.05 dB / 0.961
0.703
28.61 dB / 0.893
0.619
24.64 dB / 0.791
✓ Certified Chen et al., CVPR 2024
🥉 DeePiCt + gradient 0.695
0.796
33.05 dB / 0.953
0.676
26.27 dB / 0.839
0.612
24.36 dB / 0.781
✓ Certified Moebel et al., Nat. Methods 2021
4 DeepDeWedge + gradient 0.666
0.756
29.96 dB / 0.916
0.636
24.6 dB / 0.789
0.607
23.34 dB / 0.744
✓ Certified Wiedemann et al., Nat. Methods 2024
5 CryoSeg + gradient 0.627
0.757
30.92 dB / 0.930
0.608
23.22 dB / 0.740
0.515
20.62 dB / 0.628
✓ Certified Lamm et al., Nat. Methods 2022
6 IMOD + gradient 0.595
0.625
23.5 dB / 0.750
0.584
23.04 dB / 0.733
0.575
22.78 dB / 0.722
✓ Certified Kremer et al., J. Struct. Biol. 1996
7 IsoNet + gradient 0.571
0.722
28.32 dB / 0.887
0.538
21.28 dB / 0.658
0.454
18.06 dB / 0.503
✓ Certified Liu et al., Nat. Commun. 2021
8 SART-ET + gradient 0.547
0.561
21.61 dB / 0.673
0.573
22.33 dB / 0.704
0.506
19.88 dB / 0.593
✓ Certified Andersen & Kak, Ultrason. Imaging 1984
9 WBP + gradient 0.452
0.451
17.62 dB / 0.481
0.475
19.03 dB / 0.551
0.429
17.23 dB / 0.462
✓ Certified Crowther et al., Proc. R. Soc. 1970

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 DiffusionET + gradient 0.842 36.21 0.974
2 ETFormer + gradient 0.813 34.05 0.961
3 DeePiCt + gradient 0.796 33.05 0.953
4 CryoSeg + gradient 0.757 30.92 0.93
5 DeepDeWedge + gradient 0.756 29.96 0.916
6 IsoNet + gradient 0.722 28.32 0.887
7 IMOD + gradient 0.625 23.5 0.75
8 SART-ET + gradient 0.561 21.61 0.673
9 WBP + gradient 0.451 17.62 0.481
Spec Ranges (5 parameters)
Parameter Min Max Unit
tilt_axis_offset -0.6 1.2 px
tilt_angle_accuracy -0.2 0.4 degpertilt
dose_induced_shrinkage -2.0 4.0 -
ctf_per_tilt_variation -0.15 0.15 um
missing_wedge 26.0 38.0 deg
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 ETFormer + gradient 0.703 28.61 0.893
2 DiffusionET + gradient 0.698 28.41 0.889
3 DeePiCt + gradient 0.676 26.27 0.839
4 DeepDeWedge + gradient 0.636 24.6 0.789
5 CryoSeg + gradient 0.608 23.22 0.74
6 IMOD + gradient 0.584 23.04 0.733
7 SART-ET + gradient 0.573 22.33 0.704
8 IsoNet + gradient 0.538 21.28 0.658
9 WBP + gradient 0.475 19.03 0.551
Spec Ranges (5 parameters)
Parameter Min Max Unit
tilt_axis_offset -0.72 1.08 px
tilt_angle_accuracy -0.24 0.36 degpertilt
dose_induced_shrinkage -2.4 3.6 -
ctf_per_tilt_variation -0.15 0.15 um
missing_wedge 25.2 37.2 deg
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 ETFormer + gradient 0.619 24.64 0.791
2 DiffusionET + gradient 0.618 24.1 0.772
3 DeePiCt + gradient 0.612 24.36 0.781
4 DeepDeWedge + gradient 0.607 23.34 0.744
5 IMOD + gradient 0.575 22.78 0.722
6 CryoSeg + gradient 0.515 20.62 0.628
7 SART-ET + gradient 0.506 19.88 0.593
8 IsoNet + gradient 0.454 18.06 0.503
9 WBP + gradient 0.429 17.23 0.462
Spec Ranges (5 parameters)
Parameter Min Max Unit
tilt_axis_offset -0.42 1.38 px
tilt_angle_accuracy -0.14 0.46 degpertilt
dose_induced_shrinkage -1.4 4.6 -
ctf_per_tilt_variation -0.15 0.15 um
missing_wedge 27.2 39.2 deg
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

Π → D

Π Projection
D Detector

Mismatch Parameters

Symbol Parameter Description Nominal Perturbed
t_a tilt_axis_offset Tilt axis offset (px) 0.0 0.6
t_a tilt_angle_accuracy Tilt angle accuracy (deg per tilt) 0.0 0.2
d_s dose_induced_shrinkage Dose-induced shrinkage (-) 0.0 2.0
c_p ctf_per_tilt_variation CTF per-tilt variation (um) 0.0 0.0
m_w missing_wedge Missing wedge (deg) 30.0 34.0

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.