Reproduce “Selective Generative Restoration for Video Deraining via Uncertainty-Aware Diffusion” 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

Yi'ang Chen, Han Li, Laibin Chang, Hui Liu, Kui Jiang, Junjun Jiang

SGVD uses uncertainty estimates to decide where video deraining should preserve observed content and where a generative prior should fill obscured regions. It derives uncertainty from disagreement between noise-perturbed predictions and trains a five-frame shared-anchor constraint for temporal consistency. On synthetic tests, uncertainty correlates with restoration error; the anchor lowers warping error, though not every perceptual metric improves.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.