Skip to main content

Lund University Publications

LUND UNIVERSITY LIBRARIES

Discriminating Against Unrealistic Interpolations in Generative Adversarial Networks

Petzka, Henning LU ; Kronvall, Ted LU and Sminchisescu, Cristian LU (2022)
Abstract
Interpolations in the latent space of deep generative models is one of the standard tools to synthesize semantically meaningful mixtures of generated samples. As the generator function is non-linear, commonly used linear interpolations in the latent space do not yield the shortest paths in the sample space, resulting in non-smooth interpolations. Recent work has therefore equipped the latent space with a suitable metric to enforce shortest paths on the manifold of generated samples. These are often, however, susceptible of veering away from the manifold of real samples, resulting in smooth but unrealistic generation that requires an additional method to assess the sample quality along paths. Generative Adversarial Networks (GANs), by... (More)
Interpolations in the latent space of deep generative models is one of the standard tools to synthesize semantically meaningful mixtures of generated samples. As the generator function is non-linear, commonly used linear interpolations in the latent space do not yield the shortest paths in the sample space, resulting in non-smooth interpolations. Recent work has therefore equipped the latent space with a suitable metric to enforce shortest paths on the manifold of generated samples. These are often, however, susceptible of veering away from the manifold of real samples, resulting in smooth but unrealistic generation that requires an additional method to assess the sample quality along paths. Generative Adversarial Networks (GANs), by construction, measure the sample quality using its discriminator network. In this paper, we establish that the discriminator can be used effectively to avoid regions of low sample quality along shortest paths. By reusing the discriminator network to modify the metric on the latent space, we propose a lightweight solution for improved interpolations in pre-trained GANs. (Less)
Please use this url to cite or link to this publication:
author
; and
organization
publishing date
type
Working paper/Preprint
publication status
published
subject
keywords
Generative adversarial network (GAN), Riemannian metric, Deep learning
publisher
arXiv.org
DOI
10.48550/arXiv.2203.01035
language
English
LU publication?
yes
id
13d82c8a-8eac-4394-b275-37a427baef29
date added to LUP
2025-03-11 11:19:20
date last changed
2025-04-14 11:56:02
@misc{13d82c8a-8eac-4394-b275-37a427baef29,
  abstract     = {{Interpolations in the latent space of deep generative models is one of the standard tools to synthesize semantically meaningful mixtures of generated samples. As the generator function is non-linear, commonly used linear interpolations in the latent space do not yield the shortest paths in the sample space, resulting in non-smooth interpolations. Recent work has therefore equipped the latent space with a suitable metric to enforce shortest paths on the manifold of generated samples. These are often, however, susceptible of veering away from the manifold of real samples, resulting in smooth but unrealistic generation that requires an additional method to assess the sample quality along paths. Generative Adversarial Networks (GANs), by construction, measure the sample quality using its discriminator network. In this paper, we establish that the discriminator can be used effectively to avoid regions of low sample quality along shortest paths. By reusing the discriminator network to modify the metric on the latent space, we propose a lightweight solution for improved interpolations in pre-trained GANs.}},
  author       = {{Petzka, Henning and Kronvall, Ted and Sminchisescu, Cristian}},
  keywords     = {{Generative adversarial network (GAN); Riemannian metric; Deep learning}},
  language     = {{eng}},
  month        = {{03}},
  note         = {{Preprint}},
  publisher    = {{arXiv.org}},
  title        = {{Discriminating Against Unrealistic Interpolations in Generative Adversarial Networks}},
  url          = {{http://dx.doi.org/10.48550/arXiv.2203.01035}},
  doi          = {{10.48550/arXiv.2203.01035}},
  year         = {{2022}},
}