HackerPapers

arXiv:2104.07636eess.IV

Image Super-Resolution via Iterative Refinement

Chitwan Saharia, Jonathan Ho, William Chan, Tim Salimans, David J. Fleet, Mohammad Norouzi

Abstract

We present SR3, an approach to image Super-Resolution via Repeated Refinement. SR3 adapts denoising diffusion probabilistic models to conditional image generation and performs super-resolution through a stochastic denoising process. Inference starts with pure Gaussian noise and iteratively refines the noisy output using a U-Net model trained on denoising at various noise levels. SR3 exhibits strong performance on super-resolution tasks at different magnification factors, on faces and natural images. We conduct human evaluation on a standard 8X face super-resolution task on CelebA-HQ, comparing with SOTA GAN methods. SR3 achieves a fool rate close to 50%, suggesting photo-realistic outputs, while GANs do not exceed a fool rate of 34%. We further show the effectiveness of SR3 in cascaded image generation, where generative models are chained with super-resolution models, yielding a competitive FID score of 11.3 on ImageNet.

arXiv comments
Journal reference
DOI
Subjects
eess.IVImage and Video Processing cs.CVComputer Vision and Pattern Recognition cs.LGMachine Learning

On Hacker News

Linked by 4 identifiable commenters in 4 comments across 2 threads, from Jul 2021 to Sep 2021. Found in 1 HN submission from Jul 2021 to Jul 2021. Open submission discussions.

Comments linking this paper, per year.

Discussed in comments2 threads

  1. Google's New AI Photo Upscaling Tech Is Jaw-Dropping 3 comments

    • They do in the paper. I'm not sure why the article doesn't. https://arxiv.org/pdf/2104.07636.pdf

      AndrewKemendoView on HNParent context

    • That is indeed a shortcoming of this article in my opinion as well. If you want a comparison to the original high res photos, there are some examples in the original paper for SR3: https://arxiv.org/pdf/2104.07636.pdf Have not had a look at the CDM paper.

      BallasView on HNParent context

    • The original paper[1] itself accidentally demonstrates how dangerous this will be. Look at the picture of the leopard and study the patter of the spots around the face. The pattern on the upscaled image is clearly different from the original. The algorithm has just generated…

      dagwView on HNParent context

  2. High Fidelity Image Generation Using Diffusion Models 1 comment

    • I see only two examples of reference images on that page, but they have quite a few more examples in the paper itself: https://arxiv.org/pdf/2104.07636.pdf The results compared with reference images pretty clearly shows that the technique isn't "magical" in any sense of the term…

      bscphilView on HNParent context

Submitted to Hacker News

Connected through shared HN threads, not inferred similarity or a recommendation.