Terahertz Imaging (THz)

Terahertz Imaging (THz)

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
🥇 THz-Former 0.795 34.5 0.940 ✓ Certified THz reconstruction transformer, 2024
🥈 THz-Net 0.744 32.5 0.905 ✓ Certified Ahi et al., 2020
🥉 PnP-SPIRAL 0.630 28.5 0.810 ✓ Certified Harmany et al., 2012
4 Wiener-THz 0.498 24.5 0.680 ✓ Certified Jepsen et al., 2011

Dataset: PWM Benchmark (4 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
🥇 THz-Former + gradient 0.712
0.779
32.4 dB / 0.947
0.709
28.14 dB / 0.884
0.647
25.73 dB / 0.824
✓ Certified THz reconstruction transformer, 2024
🥈 THz-Net + gradient 0.647
0.751
30.61 dB / 0.926
0.631
24.3 dB / 0.779
0.558
22.28 dB / 0.702
✓ Certified Ahi et al., Opt. Express 2020
🥉 PnP-SPIRAL + gradient 0.618
0.667
25.63 dB / 0.821
0.590
23.47 dB / 0.749
0.596
23.37 dB / 0.745
✓ Certified Harmany et al., IEEE TCI 2012
4 Wiener-THz + gradient 0.523
0.570
21.78 dB / 0.680
0.520
20.09 dB / 0.603
0.479
19.52 dB / 0.575
✓ Certified Jepsen et al., Laser Photon. Rev. 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 THz-Former + gradient 0.779 32.4 0.947
2 THz-Net + gradient 0.751 30.61 0.926
3 PnP-SPIRAL + gradient 0.667 25.63 0.821
4 Wiener-THz + gradient 0.570 21.78 0.68
Spec Ranges (4 parameters)
Parameter Min Max Unit
pulse_chirp -0.02 0.04 ps^2
water_vapor_absorption 0.06 0.18 1/cm
beam_alignment_error -0.2 0.4 mm
dynamic_range_drift 0.98 1.04 -
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 THz-Former + gradient 0.709 28.14 0.884
2 THz-Net + gradient 0.631 24.3 0.779
3 PnP-SPIRAL + gradient 0.590 23.47 0.749
4 Wiener-THz + gradient 0.520 20.09 0.603
Spec Ranges (4 parameters)
Parameter Min Max Unit
pulse_chirp -0.024 0.036 ps^2
water_vapor_absorption 0.052 0.172 1/cm
beam_alignment_error -0.24 0.36 mm
dynamic_range_drift 0.976 1.036 -
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 THz-Former + gradient 0.647 25.73 0.824
2 PnP-SPIRAL + gradient 0.596 23.37 0.745
3 THz-Net + gradient 0.558 22.28 0.702
4 Wiener-THz + gradient 0.479 19.52 0.575
Spec Ranges (4 parameters)
Parameter Min Max Unit
pulse_chirp -0.014 0.046 ps^2
water_vapor_absorption 0.072 0.192 1/cm
beam_alignment_error -0.14 0.46 mm
dynamic_range_drift 0.986 1.046 -
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

P → D

P Propagation
D Detector

Mismatch Parameters

Symbol Parameter Description Nominal Perturbed
p_c pulse_chirp Pulse chirp (ps^2) 0.0 0.02
w_v water_vapor_absorption Water vapor absorption (1/cm) 0.1 0.14
b_a beam_alignment_error Beam alignment error (mm) 0.0 0.2
d_r dynamic_range_drift Dynamic range drift (-) 1.0 1.02

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.