[R] Diffusion Models Beat GANs on Image Synthesis

This page summarizes the projects mentioned and recommended in the original post on reddit.com/r/MachineLearning Post date: 11 May 2021

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 sample quality 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 3.85 on ImageNet $512 \times 512$. We release our code at this https URL