Wasserstein Loss for Semantic Editing in the Latent Space of GANs - Archive ouverte HAL Access content directly
Conference Papers Year : 2023

Wasserstein Loss for Semantic Editing in the Latent Space of GANs

Abstract

The latent space of GANs contains rich semantics reflecting the training data. Different methods propose to learn edits in latent space corresponding to semantic attributes, thus allowing to modify generated images. Most supervised methods rely on the guidance of classifiers to produce such edits. However, classifiers can lead to out-of-distribution regions and be fooled by adversarial samples. We propose an alternative formulation based on the Wasserstein loss that avoids such problems, while maintaining performance on-par with classifier-based approaches. We demonstrate the effectiveness of our method on two datasets (digits and faces) using StyleGAN2.
Fichier principal
Vignette du fichier
main.pdf (7.64 Mo) Télécharger le fichier
Origin Files produced by the author(s)

Dates and versions

hal-04036414 , version 1 (21-03-2023)

Identifiers

Cite

Perla Doubinsky, Nicolas Audebert, Michel Crucianu, Hervé Le Borgne. Wasserstein Loss for Semantic Editing in the Latent Space of GANs. 20th International Conference on Content-based Multimedia Indexing, Sep 2023, Orléans, France. ⟨hal-04036414⟩
158 View
39 Download

Altmetric

Share

Gmail Mastodon Facebook X LinkedIn More