Flash LiDAR

Flash LiDAR

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
🥇 DiffLiDAR 0.884 39.4 0.955 ✓ Certified Gao et al. 2024
🥈 PhysLiDAR 0.855 38.0 0.943 ✓ Certified Chen et al. 2024
🥉 SwinLiDAR 0.832 36.9 0.933 ✓ Certified Wang et al. 2023
4 TransLiDAR 0.796 35.3 0.916 ✓ Certified Li et al. 2022
5 SPADnet 0.736 32.8 0.878 ✓ Certified Lindell et al. 2018
6 DnCNN-LiDAR 0.672 30.1 0.840 ✓ Certified Peng et al. 2020
7 NL-Means-LiDAR 0.598 27.2 0.789 ✓ Certified Rapp & Goyal 2017
8 Coates-Hist 0.532 24.5 0.748 ✓ Certified Coates 1968
9 MLE-SPAD 0.489 22.8 0.718 ✓ Certified Kirmani et al. 2014

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
🥇 DiffLiDAR + gradient 0.784
0.839
36.83 dB / 0.977
0.775
33.0 dB / 0.953
0.738
31.23 dB / 0.934
✓ Certified Gao et al., NeurIPS 2024
🥈 PhysLiDAR + gradient 0.776
0.844
36.7 dB / 0.977
0.759
31.02 dB / 0.931
0.726
30.43 dB / 0.923
✓ Certified Chen et al., CVPR 2024
🥉 SwinLiDAR + gradient 0.762
0.806
34.0 dB / 0.961
0.755
31.06 dB / 0.932
0.725
29.63 dB / 0.911
✓ Certified Wang et al., ICCV 2023
4 TransLiDAR + gradient 0.746
0.788
32.96 dB / 0.952
0.748
30.48 dB / 0.924
0.703
29.18 dB / 0.903
✓ Certified Li et al., CVPR 2022
5 NL-Means-LiDAR + gradient 0.646
0.675
25.81 dB / 0.827
0.641
25.46 dB / 0.816
0.621
24.82 dB / 0.796
✓ Certified Rapp & Goyal, IEEE TCI 2017
6 SPADnet + gradient 0.608
0.773
31.04 dB / 0.931
0.551
21.3 dB / 0.659
0.499
19.97 dB / 0.597
✓ Certified Lindell et al., SIGGRAPH 2018
7 DnCNN-LiDAR + gradient 0.563
0.704
27.55 dB / 0.871
0.537
21.37 dB / 0.662
0.449
18.63 dB / 0.531
✓ Certified Peng et al., ECCV 2020
8 Coates-Hist + gradient 0.547
0.612
23.16 dB / 0.737
0.551
21.61 dB / 0.673
0.477
19.49 dB / 0.574
✓ Certified Coates, J. Phys. E 1968
9 MLE-SPAD + gradient 0.469
0.528
20.36 dB / 0.616
0.483
19.25 dB / 0.562
0.397
16.34 dB / 0.418
✓ Certified Kirmani et al., Science 2014

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 PhysLiDAR + gradient 0.844 36.7 0.977
2 DiffLiDAR + gradient 0.839 36.83 0.977
3 SwinLiDAR + gradient 0.806 34.0 0.961
4 TransLiDAR + gradient 0.788 32.96 0.952
5 SPADnet + gradient 0.773 31.04 0.931
6 DnCNN-LiDAR + gradient 0.704 27.55 0.871
7 NL-Means-LiDAR + gradient 0.675 25.81 0.827
8 Coates-Hist + gradient 0.612 23.16 0.737
9 MLE-SPAD + gradient 0.528 20.36 0.616
Spec Ranges (4 parameters)
Parameter Min Max Unit
spad_jitter -20.0 40.0 ps
ambient_photon_rate -2.0 4.0 -
pile_up_distortion -4.0 8.0 athighflux
pixel_cross_talk -1.0 2.0 -
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 DiffLiDAR + gradient 0.775 33.0 0.953
2 PhysLiDAR + gradient 0.759 31.02 0.931
3 SwinLiDAR + gradient 0.755 31.06 0.932
4 TransLiDAR + gradient 0.748 30.48 0.924
5 NL-Means-LiDAR + gradient 0.641 25.46 0.816
6 SPADnet + gradient 0.551 21.3 0.659
7 Coates-Hist + gradient 0.551 21.61 0.673
8 DnCNN-LiDAR + gradient 0.537 21.37 0.662
9 MLE-SPAD + gradient 0.483 19.25 0.562
Spec Ranges (4 parameters)
Parameter Min Max Unit
spad_jitter -24.0 36.0 ps
ambient_photon_rate -2.4 3.6 -
pile_up_distortion -4.8 7.2 athighflux
pixel_cross_talk -1.2 1.8 -
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 DiffLiDAR + gradient 0.738 31.23 0.934
2 PhysLiDAR + gradient 0.726 30.43 0.923
3 SwinLiDAR + gradient 0.725 29.63 0.911
4 TransLiDAR + gradient 0.703 29.18 0.903
5 NL-Means-LiDAR + gradient 0.621 24.82 0.796
6 SPADnet + gradient 0.499 19.97 0.597
7 Coates-Hist + gradient 0.477 19.49 0.574
8 DnCNN-LiDAR + gradient 0.449 18.63 0.531
9 MLE-SPAD + gradient 0.397 16.34 0.418
Spec Ranges (4 parameters)
Parameter Min Max Unit
spad_jitter -14.0 46.0 ps
ambient_photon_rate -1.4 4.6 -
pile_up_distortion -2.8 9.2 athighflux
pixel_cross_talk -0.7 2.3 -
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
s_j spad_jitter SPAD jitter (ps) 0.0 20.0
a_p ambient_photon_rate Ambient photon rate (-) 0.0 2.0
p_d pile_up_distortion Pile-up distortion (at high flux) 0.0 4.0
p_c pixel_cross_talk Pixel cross-talk (-) 0.0 1.0

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.