Reproduce “bit2bit: 1-bit quanta video reconstruction via self-supervised photon prediction” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Yehe Liu, Alexander Krull, Hector Basevi, Ales Leonardis, Michael W. Jenkins
bit2bit reconstructs binary SPAD video at original time steps using spatial and temporal context across frame stacks. It adapts self-supervised photon splitting with a masked loss excluding input-positive locations to address one-bit detection’s forced complementarity. On simulated data at 0.0590 photons per pixel, it achieved mean PSNR 33.93 versus 20.89 without masking. Real-data evidence is qualitative.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.