Deep feature consistent variational autoencoder

Xianxu Hou, Linlin Shen, Ke Sun, Guoping Qiu

Research output: Chapter in Book/Conference proceedingConference contributionpeer-review

138 Citations (Scopus)

Abstract

We present a novel method for constructing Variational Autoencoder (VAE). Instead of using pixel-by-pixel loss, we enforce deep feature consistency between the input and the output of a VAE, which ensures the VAE's output to preserve the spatial correlation characteristics of the input, thus leading the output to have a more natural visual appearance and better perceptual quality. Based on recent deep learning works such as style transfer, we employ a pre-Trained deep convolutional neural network (CNN) and use its hidden features to define a feature perceptual loss for VAE training. Evaluated on the CelebA face dataset, we show that our model produces better results than other methods in the literature. We also show that our method can produce latent vectors that can capture the semantic information of face expressions and can be used to achieve state-of-The-Art performance in facial attribute prediction.

Original languageEnglish
Title of host publicationProceedings - 2017 IEEE Winter Conference on Applications of Computer Vision, WACV 2017
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages1133-1141
Number of pages9
ISBN (Electronic)9781509048229
DOIs
Publication statusPublished - 11 May 2017
Event17th IEEE Winter Conference on Applications of Computer Vision, WACV 2017 - Santa Rosa, United States
Duration: 24 Mar 201731 Mar 2017

Publication series

NameProceedings - 2017 IEEE Winter Conference on Applications of Computer Vision, WACV 2017

Conference

Conference17th IEEE Winter Conference on Applications of Computer Vision, WACV 2017
Country/TerritoryUnited States
CitySanta Rosa
Period24/03/1731/03/17

ASJC Scopus subject areas

  • Computer Science Applications
  • Computer Vision and Pattern Recognition

Fingerprint

Dive into the research topics of 'Deep feature consistent variational autoencoder'. Together they form a unique fingerprint.

Cite this