Welcome to the Single Photon Challenge website! Here you'll find large synthetic single photon datasets that consist of bursts of single photon camera images and
their corresponding ground truth reconstructions. There is also a public
leaderboard showcasing various user-submitted reconstructions for this benchmark. You can directly compare submissions as well!
Single-photon cameras are capable of detecting individual photons, making them extremely sensitive,
high-speed cameras. However, each image they capture is extremely noisy and binary valued, being dominated
by photon shot noise. Given a burst of such binary frames, how can we recover a clean, high-fidelity image?
Ongoing Competition 📢
Reconstruction Competition: Consider participating in the ongoing reconstruction competition for your chance
to win thousands of dollars in prizes! Up to date competition guidelines can be
found here.
News
February 20, 2026: We identified a bug in the evaluation code that impacted the 5% and 1% metrics. As a result,
submissions submitted before 2/20/26 have had these metrics removed. We encourage participants to resubmit their
results for evaluation and, if desired, delete affected entries. Please note that missing metrics will not
influence competition judging.
February 19, 2026: New Feature! You can now click on the magnifying glass icon on a submission to see a
zoomed-in side-by-side comparison.
December 1, 2025: Fixed an issue that prevented some users from uploading their submissions from Windows
machines. We thank early testers for letting us know about it.
October 10, 2025: The dataset is now available here! Check out the FAQ for instructions on how to submit.
October 8, 2025: Hello world – this website is now live! The dataset is coming soon.
Citation
If you make use of our dataset or benchmark results, please cite:
@software{visionsim,
author = {Jungerman, Sacha and Gupta, Shantanu and Sadekar, Kaustubh and Leblang, Max and Gupta, Mohit},
license = {MIT},
month = may,
title = {{visionsim}},
url = {https://github.com/WISION-Lab/visionsim},
year = {2025}
}