Reproduce “Blind Image Restoration via Fast Diffusion Inversion” 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
Hamadi Chihaoui, Abdelhak Lemkhenter, Paolo Favaro
BIRD addresses blind image restoration by jointly optimizing the starting noise of a fixed deterministic DDIM generator and unknown degradation parameters. It generates candidate clean images and matches their degraded versions to observations. On ImageNet 4× super-resolution, it reports PSNR 22.15 and LPIPS 0.354, versus BlindDPS’s 21.36 and 0.374, but BlindDPS leads on Gaussian deblurring; BIRD assumes the degradation form is known.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.