We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:


Current browse context:


Change to browse by:

References & Citations

DBLP - CS Bibliography


(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Machine Learning

Title: Diffusion Models Beat GANs on Image Synthesis

Abstract: We show that diffusion models can achieve image sample quality superior to the current state-of-the-art generative models. We achieve this on unconditional image synthesis by finding a better architecture through a series of ablations. For conditional image synthesis, we further improve sample quality with classifier guidance: a simple, compute-efficient method for trading off diversity for fidelity using gradients from a classifier. We achieve an FID of 2.97 on ImageNet 128$\times$128, 4.59 on ImageNet 256$\times$256, and 7.72 on ImageNet 512$\times$512, and we match BigGAN-deep even with as few as 25 forward passes per sample, all while maintaining better coverage of the distribution. Finally, we find that classifier guidance combines well with upsampling diffusion models, further improving FID to 3.94 on ImageNet 256$\times$256 and 3.85 on ImageNet 512$\times$512. We release our code at this https URL
Comments: Added compute requirements, ImageNet 256$\times$256 upsampling FID and samples, DDIM guided sampler, fixed typos
Subjects: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (stat.ML)
Cite as: arXiv:2105.05233 [cs.LG]
  (or arXiv:2105.05233v4 [cs.LG] for this version)

Submission history

From: Prafulla Dhariwal [view email]
[v1] Tue, 11 May 2021 17:50:24 GMT (30977kb,D)
[v2] Wed, 12 May 2021 17:57:59 GMT (30978kb,D)
[v3] Thu, 13 May 2021 17:57:08 GMT (30980kb,D)
[v4] Tue, 1 Jun 2021 17:49:49 GMT (44152kb,D)

Link back to: arXiv, form interface, contact.