Latent reweighting, an almost free improvement for GANs - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2022

Latent reweighting, an almost free improvement for GANs

Ugo Tanielian
  • Fonction : Auteur
David Picard
Jeremie Mary

Résumé

Standard formulations of GANs, where a continuous function deforms a connected latent space, have been shown to be misspecified when fitting different classes of images. In particular, the generator will necessarily sample some lowquality images in between the classes. Rather than modifying the architecture, a line of works aims at improving the sampling quality from pre-trained generators at the expense of increased computational cost. Building on this, we introduce an additional network to predict latent importance weights and two associated sampling methods to avoid the poorest samples. This idea has several advantages: 1) it provides a way to inject disconnectedness into any GAN architecture, 2) since the rejection happens in the latent space, it avoids going through both the generator and the discriminator, saving computation time, 3) this importance weights formulation provides a principled way to reduce the Wasserstein's distance to the target distribution. We demonstrate the effectiveness of our method on several datasets, both synthetic and high-dimensional.
Fichier principal
Vignette du fichier
Issenhuth_Latent_Reweighting_an_Almost_Free_Improvement_for_GANs_WACV_2022_paper.pdf (4.01 Mo) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-03791961 , version 1 (29-09-2022)

Identifiants

Citer

Thibaut Issenhuth, Ugo Tanielian, David Picard, Jeremie Mary. Latent reweighting, an almost free improvement for GANs. 2022 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), Jan 2022, Waikoloa, United States. pp.3574-3583, ⟨10.1109/WACV51458.2022.00363⟩. ⟨hal-03791961⟩
15 Consultations
15 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More