Reproduce “GLOW-FM: Efficient Generative Low-Light Image Enhancement with One-Step Latent Flow Matching” 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
Ariel Lapid, Yuval Haitman, Reuven Peretz, Arnon Netzer, Idit Diamant
GLOW-FM performs low-light image enhancement by learning a direct transition from dark to normal-light images in the latent space of frozen Tiny-VAE. Its Terminal Velocity Matching objective predicts a finite transition, enabling one-step inference. On LSD-DLL, its small variant achieves 19.7 dB PSNR with 1 NFE and 34 GFLOPs, compared with PyDiff’s 17.3 dB, 4 NFEs, and 473 GFLOPs; this comparison does not establish lower elapsed time or superiority on every dataset or metric.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.