Event Camera / Dynamic Vision Sensor (DVS)

Event Camera / Dynamic Vision Sensor (DVS)

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
🥇 DiffEvent 0.884 39.4 0.955 ✓ Certified Gao et al. 2024
🥈 PhysEvent 0.855 38.0 0.944 ✓ Certified Chen et al. 2024
🥉 SwinEvent 0.832 36.9 0.933 ✓ Certified Zhang et al. 2023
4 TransEvent 0.794 35.2 0.914 ✓ Certified Weng et al. 2022
5 SPADE-E2VID 0.736 32.8 0.878 ✓ Certified Cadena et al. 2021
6 FireNet 0.678 30.4 0.843 ✓ Certified Scheerlinck et al. 2020
7 E2VID 0.614 27.9 0.798 ✓ Certified Rebecq et al. 2020
8 Complementary 0.537 24.8 0.748 ✓ Certified Scheerlinck et al. 2018
9 Event-Integration 0.469 22.1 0.702 ✓ Certified Mead & Mahowald 1989

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
🥇 DiffEvent + gradient 0.790
0.838
36.4 dB / 0.975
0.776
33.47 dB / 0.957
0.755
32.11 dB / 0.944
✓ Certified Gao et al., NeurIPS 2024
🥈 PhysEvent + gradient 0.755
0.823
35.93 dB / 0.973
0.735
29.37 dB / 0.907
0.706
28.43 dB / 0.890
✓ Certified Chen et al., ECCV 2024
🥉 SwinEvent + gradient 0.745
0.808
34.11 dB / 0.962
0.761
31.6 dB / 0.938
0.666
26.13 dB / 0.836
✓ Certified Zhang et al., CVPR 2023
4 TransEvent + gradient 0.721
0.785
32.73 dB / 0.950
0.714
29.0 dB / 0.900
0.664
26.35 dB / 0.842
✓ Certified Weng et al., ECCV 2022
5 SPADE-E2VID + gradient 0.609
0.749
30.14 dB / 0.919
0.581
22.78 dB / 0.722
0.497
19.36 dB / 0.568
✓ Certified Cadena et al., IEEE TIP 2021
6 Complementary + gradient 0.582
0.614
23.06 dB / 0.733
0.596
23.23 dB / 0.740
0.537
21.02 dB / 0.647
✓ Certified Scheerlinck et al., RA-L 2018
7 FireNet + gradient 0.569
0.709
27.94 dB / 0.880
0.517
20.25 dB / 0.611
0.480
19.28 dB / 0.564
✓ Certified Scheerlinck et al., WACV 2020
8 E2VID + gradient 0.522
0.664
25.78 dB / 0.826
0.481
18.97 dB / 0.548
0.422
17.66 dB / 0.483
✓ Certified Rebecq et al., IEEE TPAMI 2020
9 Event-Integration + gradient 0.497
0.502
19.4 dB / 0.570
0.517
20.26 dB / 0.611
0.471
19.1 dB / 0.555
✓ Certified Mead & Mahowald, Analog VLSI 1989

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 DiffEvent + gradient 0.838 36.4 0.975
2 PhysEvent + gradient 0.823 35.93 0.973
3 SwinEvent + gradient 0.808 34.11 0.962
4 TransEvent + gradient 0.785 32.73 0.95
5 SPADE-E2VID + gradient 0.749 30.14 0.919
6 FireNet + gradient 0.709 27.94 0.88
7 E2VID + gradient 0.664 25.78 0.826
8 Complementary + gradient 0.614 23.06 0.733
9 Event-Integration + gradient 0.502 19.4 0.57
Spec Ranges (4 parameters)
Parameter Min Max Unit
contrast_threshold 0.26 0.38 logintensity
refractory_period -0.8 4.6 us
noise_event_rate -0.2 0.4 ofrealevents
hot_pixel_fraction -0.1 0.2 -
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 DiffEvent + gradient 0.776 33.47 0.957
2 SwinEvent + gradient 0.761 31.6 0.938
3 PhysEvent + gradient 0.735 29.37 0.907
4 TransEvent + gradient 0.714 29.0 0.9
5 Complementary + gradient 0.596 23.23 0.74
6 SPADE-E2VID + gradient 0.581 22.78 0.722
7 FireNet + gradient 0.517 20.25 0.611
8 Event-Integration + gradient 0.517 20.26 0.611
9 E2VID + gradient 0.481 18.97 0.548
Spec Ranges (4 parameters)
Parameter Min Max Unit
contrast_threshold 0.252 0.372 logintensity
refractory_period -1.16 4.24 us
noise_event_rate -0.24 0.36 ofrealevents
hot_pixel_fraction -0.12 0.18 -
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 DiffEvent + gradient 0.755 32.11 0.944
2 PhysEvent + gradient 0.706 28.43 0.89
3 SwinEvent + gradient 0.666 26.13 0.836
4 TransEvent + gradient 0.664 26.35 0.842
5 Complementary + gradient 0.537 21.02 0.647
6 SPADE-E2VID + gradient 0.497 19.36 0.568
7 FireNet + gradient 0.480 19.28 0.564
8 Event-Integration + gradient 0.471 19.1 0.555
9 E2VID + gradient 0.422 17.66 0.483
Spec Ranges (4 parameters)
Parameter Min Max Unit
contrast_threshold 0.272 0.392 logintensity
refractory_period -0.26 5.14 us
noise_event_rate -0.14 0.46 ofrealevents
hot_pixel_fraction -0.07 0.23 -
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 → D

M Modulation
D Detector

Mismatch Parameters

Symbol Parameter Description Nominal Perturbed
c_t contrast_threshold Contrast threshold (log intensity) 0.3 0.34
r_p refractory_period Refractory period (us) 1.0 2.8
n_e noise_event_rate Noise event rate (of real events) 0.0 0.2
h_p hot_pixel_fraction Hot pixel fraction (-) 0.0 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.