Reproduce “Steered Diffusion: A Generalized Framework for Plug-and-Play Conditional Image Synthesis” 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
Nithin Gopalakrishnan Nair, Anoop Cherian, Suhas Lohit, Ye Wang, Toshiaki Koike-Akino, Vishal M. Patel, Tim K. Marks
Steered Diffusion repurposes an unconditional image generator for conditional tasks without task-specific retraining. It evaluates a rough clean-image prediction with an inverse model or known degradation, then steers denoising toward the condition; some tasks use repeated refinement. On 300 CelebA images, it achieves 26.02 dB PSNR versus ILVR’s 20.54 dB for 8× face super-resolution, not evidence of universal restoration quality.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.