Brillouin Microscopy

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
🥇	DiffusionSpectra DiffusionSpectra Gao et al., Nat. Methods 2024 39.5 dB SSIM 0.963 Checkpoint unavailable	0.890	39.5	0.963	✓ Certified	Gao et al., Nat. Methods 2024
🥈	SpectraFormer SpectraFormer Chen et al., arXiv 2023 38.4 dB SSIM 0.954 Checkpoint unavailable	0.867	38.4	0.954	✓ Certified	Chen et al., arXiv 2023
🥉	PINN-Brillouin PINN-Brillouin Raissi et al., J. Comput. Phys. 2019 (adapted) 37.0 dB SSIM 0.942 Checkpoint unavailable	0.838	37.0	0.942	✓ Certified	Raissi et al., J. Comput. Phys. 2019 (adapted)
4	U-Net-Spectral U-Net-Spectral Ronneberger et al., MICCAI 2015 (spectral) 36.1 dB SSIM 0.933 Checkpoint unavailable	0.818	36.1	0.933	✓ Certified	Ronneberger et al., MICCAI 2015 (spectral)
5	CDAE CDAE Zhang et al., Sensors 2024 34.8 dB SSIM 0.918 Checkpoint unavailable	0.789	34.8	0.918	✓ Certified	Zhang et al., Sensors 2024
6	DnCNN-Brillouin DnCNN-Brillouin Zhang et al., IEEE TIP 2017 (adapted) 33.2 dB SSIM 0.901 Checkpoint unavailable	0.754	33.2	0.901	✓ Certified	Zhang et al., IEEE TIP 2017 (adapted)
7	CNN-Spectra CNN-Spectra Remer & Bhatt, Biomed. Opt. Express 2020 31.5 dB SSIM 0.872 Checkpoint unavailable	0.711	31.5	0.872	✓ Certified	Remer & Bhatt, Biomed. Opt. Express 2020
8	SG-Baseline SG-Baseline Savitzky & Golay, Anal. Chem. 1964 27.8 dB SSIM 0.812 Try in SpecLab →	0.619	27.8	0.812	✓ Certified	Savitzky & Golay, Anal. Chem. 1964
9	Lorentzian-Fit Lorentzian-Fit Dil, Rep. Prog. Phys. 1982 26.2 dB SSIM 0.785 Try in SpecLab →	0.579	26.2	0.785	✓ Certified	Dil, Rep. Prog. Phys. 1982

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
🥇	SpectraFormer + gradient SpectraFormer + gradient Chen et al., arXiv 2023 Score 0.814 Correct & Reconstruct →	0.814	0.848 36.69 dB / 0.977	0.817 35.81 dB / 0.972	0.777 32.42 dB / 0.947	✓ Certified	Chen et al., arXiv 2023
🥈	DiffusionSpectra + gradient DiffusionSpectra + gradient Gao et al., Nat. Methods 2024 Score 0.777 Correct & Reconstruct →	0.777	0.839 36.85 dB / 0.977	0.756 31.1 dB / 0.932	0.736 30.83 dB / 0.929	✓ Certified	Gao et al., Nat. Methods 2024
🥉	PINN-Brillouin + gradient PINN-Brillouin + gradient Raissi et al., J. Comput. Phys. 2019 (adapted) Score 0.745 Correct & Reconstruct →	0.745	0.808 34.29 dB / 0.963	0.736 30.47 dB / 0.924	0.692 28.21 dB / 0.885	✓ Certified	Raissi et al., J. Comput. Phys. 2019 (adapted)
4	CDAE + gradient CDAE + gradient Zhang et al., Sensors 2024 Score 0.726 Correct & Reconstruct →	0.726	0.780 32.35 dB / 0.946	0.700 27.68 dB / 0.874	0.699 28.17 dB / 0.884	✓ Certified	Zhang et al., Sensors 2024
5	U-Net-Spectral + gradient U-Net-Spectral + gradient Ronneberger et al., MICCAI 2015 (spectral) Score 0.719 Correct & Reconstruct →	0.719	0.799 33.93 dB / 0.960	0.693 27.17 dB / 0.862	0.664 26.43 dB / 0.844	✓ Certified	Ronneberger et al., MICCAI 2015 (spectral)
6	SG-Baseline + gradient SG-Baseline + gradient Savitzky & Golay, Anal. Chem. 1964 Score 0.637 Correct & Reconstruct →	0.637	0.662 25.7 dB / 0.823	0.642 24.79 dB / 0.795	0.608 23.71 dB / 0.758	✓ Certified	Savitzky & Golay, Anal. Chem. 1964
7	DnCNN-Brillouin + gradient DnCNN-Brillouin + gradient Zhang et al., IEEE TIP 2017 (adapted) Score 0.632 Correct & Reconstruct →	0.632	0.757 30.61 dB / 0.926	0.610 24.1 dB / 0.772	0.530 21.42 dB / 0.665	✓ Certified	Zhang et al., IEEE TIP 2017 (adapted)
8	CNN-Spectra + gradient CNN-Spectra + gradient Remer & Bhatt, Biomed. Opt. Express 2020 Score 0.600 Correct & Reconstruct →	0.600	0.758 30.46 dB / 0.924	0.570 22.0 dB / 0.690	0.473 18.83 dB / 0.541	✓ Certified	Remer & Bhatt, Biomed. Opt. Express 2020
9	Lorentzian-Fit + gradient Lorentzian-Fit + gradient Dil, Rep. Prog. Phys. 1982 Score 0.593 Correct & Reconstruct →	0.593	0.614 23.47 dB / 0.749	0.598 22.99 dB / 0.731	0.566 22.74 dB / 0.721	✓ Certified	Dil, Rep. Prog. Phys. 1982

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	SpectraFormer + gradient	0.848	36.69	0.977
2	DiffusionSpectra + gradient	0.839	36.85	0.977
3	PINN-Brillouin + gradient	0.808	34.29	0.963
4	U-Net-Spectral + gradient	0.799	33.93	0.96
5	CDAE + gradient	0.780	32.35	0.946
6	CNN-Spectra + gradient	0.758	30.46	0.924
7	DnCNN-Brillouin + gradient	0.757	30.61	0.926
8	SG-Baseline + gradient	0.662	25.7	0.823
9	Lorentzian-Fit + gradient	0.614	23.47	0.749

Spec Ranges (3 parameters)

Parameter	Min	Max	Unit
brillouin_shift_calibration	-10.0	20.0	MHz
vipa_fsr_error	-0.1	0.2	-
elastic_scattering_leakage	-12.0	6.0	-

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	SpectraFormer + gradient	0.817	35.81	0.972
2	DiffusionSpectra + gradient	0.756	31.1	0.932
3	PINN-Brillouin + gradient	0.736	30.47	0.924
4	CDAE + gradient	0.700	27.68	0.874
5	U-Net-Spectral + gradient	0.693	27.17	0.862
6	SG-Baseline + gradient	0.642	24.79	0.795
7	DnCNN-Brillouin + gradient	0.610	24.1	0.772
8	Lorentzian-Fit + gradient	0.598	22.99	0.731
9	CNN-Spectra + gradient	0.570	22.0	0.69

Spec Ranges (3 parameters)

Parameter	Min	Max	Unit
brillouin_shift_calibration	-12.0	18.0	MHz
vipa_fsr_error	-0.12	0.18	-
elastic_scattering_leakage	-10.8	7.2	-

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	SpectraFormer + gradient	0.777	32.42	0.947
2	DiffusionSpectra + gradient	0.736	30.83	0.929
3	CDAE + gradient	0.699	28.17	0.884
4	PINN-Brillouin + gradient	0.692	28.21	0.885
5	U-Net-Spectral + gradient	0.664	26.43	0.844
6	SG-Baseline + gradient	0.608	23.71	0.758
7	Lorentzian-Fit + gradient	0.566	22.74	0.721
8	DnCNN-Brillouin + gradient	0.530	21.42	0.665
9	CNN-Spectra + gradient	0.473	18.83	0.541

Spec Ranges (3 parameters)

Parameter	Min	Max	Unit
brillouin_shift_calibration	-7.0	23.0	MHz
vipa_fsr_error	-0.07	0.23	-
elastic_scattering_leakage	-13.8	4.2	-

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	Perturbed
b_s	brillouin_shift_calibration	Brillouin shift calibration (MHz)	10.0
v_f	vipa_fsr_error	VIPA FSR error (-)	0.1
e_s	elastic_scattering_leakage	Elastic scattering leakage (-)	-6.0

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.