Ocean Acoustic Tomography

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
🥇	DiffusionExperimental DiffusionExperimental Zhang et al., 2024 34.23 dB SSIM 0.963 Checkpoint unavailable	0.802	34.23	0.963	✓ Certified	Zhang et al., 2024
🥈	ScoreExperimental ScoreExperimental Wei et al., 2025 33.76 dB SSIM 0.959 Checkpoint unavailable	0.792	33.76	0.959	✓ Certified	Wei et al., 2025
🥉	SwinIR SwinIR Liang et al., ICCVW 2021 34.1 dB SSIM 0.942 Checkpoint unavailable	0.789	34.1	0.942	✓ Certified	Liang et al., ICCVW 2021
4	Domain-Adapted-CNN Domain-Adapted-CNN Domain adaptation CNN 33.58 dB SSIM 0.958 Checkpoint unavailable	0.789	33.58	0.958	✓ Certified	Domain adaptation CNN
5	ExpFormer ExpFormer Experimental science transformer, 2024 31.88 dB SSIM 0.941 Checkpoint unavailable	0.752	31.88	0.941	✓ Certified	Experimental science transformer, 2024
6	ResUNet ResUNet Residual U-Net baseline 32.6 dB SSIM 0.915 Checkpoint unavailable	0.751	32.6	0.915	✓ Certified	Residual U-Net baseline
7	PnP-ADMM PnP-ADMM ADMM + denoiser prior 29.4 dB SSIM 0.907 Try in SpecLab →	0.694	29.4	0.907	✓ Certified	ADMM + denoiser prior
8	PnP-RED PnP-RED Romano et al., IEEE TIP 2017 28.9 dB SSIM 0.835 Try in SpecLab →	0.649	28.9	0.835	✓ Certified	Romano et al., IEEE TIP 2017
9	Matched Filter Matched Filter Optimal linear filter 25.7 dB SSIM 0.823 Try in SpecLab →	0.590	25.7	0.823	✓ Certified	Optimal linear filter
10	Wiener Filter Wiener Filter Wiener filtering baseline 25.62 dB SSIM 0.821 Try in SpecLab →	0.588	25.62	0.821	✓ Certified	Wiener filtering baseline
11	Tikhonov Tikhonov Analytical baseline 25.4 dB SSIM 0.710 Try in SpecLab →	0.528	25.4	0.710	✓ Certified	Analytical baseline

Dataset: PWM Benchmark (11 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
🥇	SwinIR + gradient SwinIR + gradient Liang et al., ICCVW 2021 Score 0.708 Correct & Reconstruct →	0.708	0.772 32.07 dB / 0.943	0.701 28.6 dB / 0.893	0.650 26.04 dB / 0.833	✓ Certified	Liang et al., ICCVW 2021
🥈	Domain-Adapted-CNN + gradient Domain-Adapted-CNN + gradient Domain adaptation CNN Score 0.698 Correct & Reconstruct →	0.698	0.767 31.69 dB / 0.939	0.693 27.84 dB / 0.877	0.633 25.53 dB / 0.818	✓ Certified	Domain adaptation CNN
🥉	ScoreExperimental + gradient ScoreExperimental + gradient Wei et al., 2025 Score 0.689 Correct & Reconstruct →	0.689	0.764 30.97 dB / 0.930	0.668 26.65 dB / 0.849	0.635 25.0 dB / 0.802	✓ Certified	Wei et al., 2025
4	DiffusionExperimental + gradient DiffusionExperimental + gradient Zhang et al., 2024 Score 0.684 Correct & Reconstruct →	0.684	0.797 33.15 dB / 0.954	0.652 25.42 dB / 0.815	0.602 23.32 dB / 0.743	✓ Certified	Zhang et al., 2024
5	ExpFormer + gradient ExpFormer + gradient Experimental science transformer, 2024 Score 0.642 Correct & Reconstruct →	0.642	0.732 28.98 dB / 0.900	0.632 24.28 dB / 0.778	0.562 22.14 dB / 0.696	✓ Certified	Experimental science transformer, 2024
6	ResUNet + gradient ResUNet + gradient Residual U-Net baseline Score 0.638 Correct & Reconstruct →	0.638	0.772 31.26 dB / 0.934	0.625 24.42 dB / 0.783	0.518 20.84 dB / 0.638	✓ Certified	Residual U-Net baseline
7	PnP-RED + gradient PnP-RED + gradient Romano et al., IEEE TIP 2017 Score 0.626 Correct & Reconstruct →	0.626	0.681 26.5 dB / 0.846	0.627 24.65 dB / 0.791	0.569 22.58 dB / 0.714	✓ Certified	Romano et al., IEEE TIP 2017
8	Tikhonov + gradient Tikhonov + gradient Tikhonov, Doklady 1963 Score 0.593 Correct & Reconstruct →	0.593	0.638 24.28 dB / 0.778	0.581 22.41 dB / 0.707	0.560 22.46 dB / 0.709	✓ Certified	Tikhonov, Doklady 1963
9	PnP-ADMM + gradient PnP-ADMM + gradient ADMM + denoiser prior Score 0.587 Correct & Reconstruct →	0.587	0.686 26.51 dB / 0.846	0.570 22.52 dB / 0.712	0.505 19.7 dB / 0.584	✓ Certified	ADMM + denoiser prior
10	Matched Filter + gradient Matched Filter + gradient Optimal linear filter Score 0.584 Correct & Reconstruct →	0.584	0.605 23.1 dB / 0.735	0.612 23.8 dB / 0.761	0.534 21.55 dB / 0.670	✓ Certified	Optimal linear filter
11	Wiener Filter + gradient Wiener Filter + gradient Wiener filtering baseline Score 0.549 Correct & Reconstruct →	0.549	0.603 23.1 dB / 0.735	0.554 21.86 dB / 0.684	0.489 20.04 dB / 0.601	✓ Certified	Wiener filtering baseline

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 5 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	DiffusionExperimental + gradient	0.797	33.15	0.954
2	SwinIR + gradient	0.772	32.07	0.943
3	ResUNet + gradient	0.772	31.26	0.934
4	Domain-Adapted-CNN + gradient	0.767	31.69	0.939
5	ScoreExperimental + gradient	0.764	30.97	0.93
6	ExpFormer + gradient	0.732	28.98	0.9
7	PnP-ADMM + gradient	0.686	26.51	0.846
8	PnP-RED + gradient	0.681	26.5	0.846
9	Tikhonov + gradient	0.638	24.28	0.778
10	Matched Filter + gradient	0.605	23.1	0.735
11	Wiener Filter + gradient	0.603	23.1	0.735

Spec Ranges (4 parameters)

Parameter	Min	Max	Unit
sound_speed_profile_error	-0.4	0.8	-
multipath_identification	-4.0	8.0	-
source/receiver_position	-2.0	4.0	m
current_velocity_error	-0.1	0.2	m/s

View Details →

Dev 5 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	SwinIR + gradient	0.701	28.6	0.893
2	Domain-Adapted-CNN + gradient	0.693	27.84	0.877
3	ScoreExperimental + gradient	0.668	26.65	0.849
4	DiffusionExperimental + gradient	0.652	25.42	0.815
5	ExpFormer + gradient	0.632	24.28	0.778
6	PnP-RED + gradient	0.627	24.65	0.791
7	ResUNet + gradient	0.625	24.42	0.783
8	Matched Filter + gradient	0.612	23.8	0.761
9	Tikhonov + gradient	0.581	22.41	0.707
10	PnP-ADMM + gradient	0.570	22.52	0.712
11	Wiener Filter + gradient	0.554	21.86	0.684

Spec Ranges (4 parameters)

Parameter	Min	Max	Unit
sound_speed_profile_error	-0.48	0.72	-
multipath_identification	-4.8	7.2	-
source/receiver_position	-2.4	3.6	m
current_velocity_error	-0.12	0.18	m/s

Submit Reconstruction View Details →

Hidden 5 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	SwinIR + gradient	0.650	26.04	0.833
2	ScoreExperimental + gradient	0.635	25.0	0.802
3	Domain-Adapted-CNN + gradient	0.633	25.53	0.818
4	DiffusionExperimental + gradient	0.602	23.32	0.743
5	PnP-RED + gradient	0.569	22.58	0.714
6	ExpFormer + gradient	0.562	22.14	0.696
7	Tikhonov + gradient	0.560	22.46	0.709
8	Matched Filter + gradient	0.534	21.55	0.67
9	ResUNet + gradient	0.518	20.84	0.638
10	PnP-ADMM + gradient	0.505	19.7	0.584
11	Wiener Filter + gradient	0.489	20.04	0.601

Spec Ranges (4 parameters)

Parameter	Min	Max	Unit
sound_speed_profile_error	-0.28	0.92	-
multipath_identification	-2.8	9.2	-
source/receiver_position	-1.4	4.6	m
current_velocity_error	-0.07	0.23	m/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	Perturbed
s_s	sound_speed_profile_error	Sound speed profile error (-)	0.4
m_i	multipath_identification	Multipath identification (-)	4.0
s_p	source/receiver_position	Source/receiver position (m)	2.0
c_v	current_velocity_error	Current velocity error (m/s)	0.1

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.