Active Thermography (IR)

Active Thermography (IR)

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
🥇 DiffusionThermo 0.817 35.5 0.950 ✓ Certified Diffusion model for thermal imaging, 2024
🥈 ThermoFormer 0.794 34.5 0.938 ✓ Certified Thermography transformer, 2024
🥉 PINN-Thermo 0.760 33.0 0.920 ✓ Certified PINN thermography extension 2024
4 U-Net Thermo 0.736 32.0 0.905 ✓ Certified Fang et al., IEEE TIM 2023
5 ThermoNet 0.685 30.0 0.870 ✓ Certified Hu et al., NDT&E Int. 2024
6 PnP-ADMM 0.595 27.0 0.790 ✓ Certified Venkatakrishnan et al., 2013
7 PCT 0.495 24.0 0.690 ✓ Certified Maldague & Marinetti, 1996
8 TSR 0.427 22.0 0.620 ✓ Certified Shepard et al., Opt. Eng. 2003

Dataset: PWM Benchmark (8 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
🥇 ThermoFormer + gradient 0.735
0.777
32.42 dB / 0.947
0.743
30.51 dB / 0.924
0.686
26.91 dB / 0.856
✓ Certified Transformer for thermography reconstruction, 2024
🥈 DiffusionThermo + gradient 0.706
0.790
33.04 dB / 0.953
0.678
26.45 dB / 0.844
0.650
25.8 dB / 0.826
✓ Certified Score-based diffusion for thermal imaging, 2024
🥉 PINN-Thermo + gradient 0.680
0.754
30.6 dB / 0.926
0.658
26.27 dB / 0.839
0.627
24.22 dB / 0.776
✓ Certified Raissi et al. 2019; thermography extension 2024
4 U-Net Thermo + gradient 0.647
0.737
29.4 dB / 0.907
0.623
24.83 dB / 0.797
0.580
22.44 dB / 0.708
✓ Certified Fang et al., IEEE Trans. Instrum. Meas. 2023
5 PnP-ADMM + gradient 0.592
0.648
25.12 dB / 0.806
0.564
22.38 dB / 0.706
0.565
22.09 dB / 0.694
✓ Certified Venkatakrishnan et al., IEEE GlobalSIP 2013
6 ThermoNet + gradient 0.577
0.701
27.5 dB / 0.870
0.568
22.35 dB / 0.705
0.463
18.56 dB / 0.528
✓ Certified Hu et al., NDT&E Int. 2024
7 PCT + gradient 0.540
0.601
22.71 dB / 0.720
0.524
20.37 dB / 0.616
0.495
19.42 dB / 0.571
✓ Certified Maldague & Marinetti, J. Appl. Phys. 1996
8 TSR + gradient 0.506
0.550
20.91 dB / 0.642
0.502
20.09 dB / 0.603
0.466
19.19 dB / 0.559
✓ Certified Shepard, Thermosense 2001; Shepard et al., Opt. Eng. 2003

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 DiffusionThermo + gradient 0.790 33.04 0.953
2 ThermoFormer + gradient 0.777 32.42 0.947
3 PINN-Thermo + gradient 0.754 30.6 0.926
4 U-Net Thermo + gradient 0.737 29.4 0.907
5 ThermoNet + gradient 0.701 27.5 0.87
6 PnP-ADMM + gradient 0.648 25.12 0.806
7 PCT + gradient 0.601 22.71 0.72
8 TSR + gradient 0.550 20.91 0.642
Spec Ranges (4 parameters)
Parameter Min Max Unit
emissivity_error 0.94 0.97 -
heat_source_power_drift 0.98 1.04 -
background_temperature 24.0 27.0 C
integration_time_offset -0.02 0.04 s
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 ThermoFormer + gradient 0.743 30.51 0.924
2 DiffusionThermo + gradient 0.678 26.45 0.844
3 PINN-Thermo + gradient 0.658 26.27 0.839
4 U-Net Thermo + gradient 0.623 24.83 0.797
5 ThermoNet + gradient 0.568 22.35 0.705
6 PnP-ADMM + gradient 0.564 22.38 0.706
7 PCT + gradient 0.524 20.37 0.616
8 TSR + gradient 0.502 20.09 0.603
Spec Ranges (4 parameters)
Parameter Min Max Unit
emissivity_error 0.938 0.968 -
heat_source_power_drift 0.976 1.036 -
background_temperature 23.8 26.8 C
integration_time_offset -0.024 0.036 s
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 ThermoFormer + gradient 0.686 26.91 0.856
2 DiffusionThermo + gradient 0.650 25.8 0.826
3 PINN-Thermo + gradient 0.627 24.22 0.776
4 U-Net Thermo + gradient 0.580 22.44 0.708
5 PnP-ADMM + gradient 0.565 22.09 0.694
6 PCT + gradient 0.495 19.42 0.571
7 TSR + gradient 0.466 19.19 0.559
8 ThermoNet + gradient 0.463 18.56 0.528
Spec Ranges (4 parameters)
Parameter Min Max Unit
emissivity_error 0.943 0.973 -
heat_source_power_drift 0.986 1.046 -
background_temperature 24.3 27.3 C
integration_time_offset -0.014 0.046 s
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
e_e emissivity_error Emissivity error (-) 0.95 0.96
h_s heat_source_power_drift Heat source power drift (-) 1.0 1.02
b_t background_temperature Background temperature (C) 25.0 26.0
i_t integration_time_offset Integration time offset (s) 0.0 0.02

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.