Disentangled latent representations of images with atomic autoencoders - Laboratoire Traitement et Communication de l'Information Access content directly
Conference Papers Year : 2023

Disentangled latent representations of images with atomic autoencoders

Abstract

We present the atomic autoencoder architecture, which decomposes an image as the sum of elementary parts that are parametrized by simple separate blocks of latent codes. We show that this simple architecture is induced by the definition of a general atomic low-dimensional model of the considered data. We also highlight the fact that the atomic autoencoder achieves disentangled low-dimensional representations under minimal hypotheses. Experiments show that their implementation with deep neural networks is successful at learning disentangled representations on two different examples: images constructed with simple parametric curves and images of filtered off-the-grid spikes.
Fichier principal
Vignette du fichier
atomic_autoencoders_hal2.pdf (351.42 Ko) Télécharger le fichier
atomic_autoencoders_hal2_supp_mat.pdf (351.19 Ko) Télécharger le fichier
Origin Files produced by the author(s)
Origin Files produced by the author(s)

Dates and versions

hal-03962759 , version 1 (30-01-2023)
hal-03962759 , version 2 (22-05-2023)

Identifiers

  • HAL Id : hal-03962759 , version 2

Cite

Alasdair Newson, Yann Traonmilin. Disentangled latent representations of images with atomic autoencoders. Sampling Theory and Applications Conference, Jul 2023, Sampling Theory and Applications Conference, United States. ⟨hal-03962759v2⟩
326 View
129 Download

Share

Gmail Mastodon Facebook X LinkedIn More