Diffusion Models Beat GANs on Image Synthesis

Code

• github.com/openai/guided-diffusion
  OfficialIn paperpytorch★ 7,321
• github.com/Janspiry/Palette-Image-to-Image-Diffusion-Models
  pytorch★ 1,807
• github.com/clu0/unet.cu
  pytorch★ 656
• github.com/afiaka87/clip-guided-diffusion
  pytorch★ 460
• github.com/ML-GSAI/EGSDE
  pytorch★ 234
• github.com/jychoi118/p2-weighting
  pytorch★ 153
• github.com/crowsonkb/guided-diffusion
  pytorch★ 150
• github.com/lucidrains/ddpm-proteins
  pytorch★ 144
• github.com/LouisRouss/Diffusion-Based-Model-for-Colorization
  pytorch★ 139
• github.com/filipbasara0/simple-diffusion
  pytorch★ 133

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 128128, 4.59 on ImageNet 256256, and 7.72 on ImageNet 512512, 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 256256 and 3.85 on ImageNet 512512. We release our code at https://github.com/openai/guided-diffusion

Tasks

Benchmark Results

Reproductions