Generative Modelling With Inverse Heat Dissipation

06/21/2022
by   Severi Rissanen, et al.
0

While diffusion models have shown great success in image generation, their noise-inverting generative process does not explicitly consider the structure of images, such as their inherent multi-scale nature. Inspired by diffusion models and the desirability of coarse-to-fine modelling, we propose a new model that generates images through iteratively inverting the heat equation, a PDE that locally erases fine-scale information when run over the 2D plane of the image. In our novel methodology, the solution of the forward heat equation is interpreted as a variational approximation in a directed graphical model. We demonstrate promising image quality and point out emergent qualitative properties not seen in diffusion models, such as disentanglement of overall colour and shape in images and aspects of neural network interpretability. Spectral analysis on natural images positions our model as a type of dual to diffusion models and reveals implicit inductive biases in them.

READ FULL TEXT

page 30

page 31

page 32

page 33

page 34

page 37

page 38

page 39

research
09/12/2022

Blurring Diffusion Models

Recently, Rissanen et al., (2022) have presented a new type of diffusion...
research
09/28/2021

An extended Krylov subspace method for decoding edge-based compressed images by homogeneous diffusion

The heat equation is often used in order to inpaint dropped data in inpa...
research
08/30/2023

Physics-Informed DeepMRI: Bridging the Gap from Heat Diffusion to k-Space Interpolation

In the field of parallel imaging (PI), alongside image-domain regulariza...
research
07/14/2023

Inverse Evolution Layers: Physics-informed Regularizers for Deep Neural Networks

This paper proposes a novel approach to integrating partial differential...
research
04/02/2023

Textile Pattern Generation Using Diffusion Models

The problem of text-guided image generation is a complex task in Compute...
research
05/05/2023

Generative Steganography Diffusion

Generative steganography (GS) is an emerging technique that generates st...
research
07/26/2023

Founding a mathematical diffusion model in linguistics. The case study of German syntactic features in the North-Eastern Italian dialects

We take as a case study the spread of Germanic syntactic features into R...

Please sign up or login with your details

Forgot password? Click here to reset