CeFET: Contrast-Enhanced Guided Facial Expression Translation

Linfeng Han, Zhong Jin, Yi Chang Li*, Zhiyang Jia

*Corresponding author for this work

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

Abstract

Advancements in deep learning have stimulated the creation of multiple techniques for facial expression translation. However, these methods frequently rely on detailed annotations of action units (AU) or 3D modelling techniques. In this paper, we introduce a novel Contrast-enhanced Guided Facial Expression Translation (CeFET) method. The model uses only facial images as input and extracts facial features from these images using an encoder model based on the StyleGAN prior. We propose a contrast-enhanced guidance technique aimed at minimizing the distance between the generated face and the input face, as well as the distance between the generated expression and the reference expression. This ensures that the generated face maintains identity consistency with the source face and expression consistency with the reference face. Extensive experimental results support the effectiveness of our method.

Original languageEnglish
Title of host publicationProceedings of the 14th International Conference on Computer Engineering and Networks - Volume IV
EditorsGuangqiang Yin, Xiaodong Liu, Jian Su, Yangzhao Yang
PublisherSpringer Science and Business Media Deutschland GmbH
Pages177-188
Number of pages12
ISBN (Print)9789819640157
DOIs
Publication statusPublished - 2025
Externally publishedYes
Event14th International Conference on Computer Engineering and Networks, CENet 2024 - Kashi, China
Duration: 18 Oct 202421 Oct 2024

Publication series

NameLecture Notes in Electrical Engineering
Volume1383 LNEE
ISSN (Print)1876-1100
ISSN (Electronic)1876-1119

Conference

Conference14th International Conference on Computer Engineering and Networks, CENet 2024
Country/TerritoryChina
CityKashi
Period18/10/2421/10/24

Keywords

  • Contrastive Learning
  • Facial Expression Translation
  • Generative Adversarial Network

Fingerprint

Dive into the research topics of 'CeFET: Contrast-Enhanced Guided Facial Expression Translation'. Together they form a unique fingerprint.

Cite this