This article discusses the concept of denoising diffusion, which is a technique used in generative AI to shape random noise into novel samples of data. It explains the theory behind denoising diffusion and explores the design choices related to sampling and training the denoiser network. The article presents key findings and insights through visualizations and code.
•19m read time• From developer.nvidia.com
Table of contents
Denoising diffusionWhat makes diffusion work?Design choices for sampling to generate imagesDesign choices for training the denoiserResults and conclusionsSort: