Deep Diffusion Models for Facies Modeling
Lukas Mosser · 2023
Summary Deep diffusion models are a recently developed approach to model distributions over images. In this presentation we will explore order agnostic diffusion models and their relationship with traditional autoregressive facies modeling approaches. We show applications on geoscience relevant datasets and training images. In addition, we present an extension to handle domains larger than the original training image size. Finally, we highlight benefits of the approach relating to the probabilistic nature of the approach.