Skip to main navigation Skip to search Skip to main content

Exposing GAN-synthesized faces using landmark locations

  • SUNY Albany
  • Chinese Academy of Sciences

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

115 Scopus citations

Abstract

Generative adversary networks (GANs) have recently led to highly realistic image synthesis results. In this work, we describe a new method to expose GAN-synthesized images using the locations of the facial landmark points. Our method is based on the observations that the facial parts configuration generated by GAN models are different from those of the real faces, due to the lack of global constraints. We perform experiments demonstrating this phenomenon, and show that an SVM classifier trained using the locations of facial landmark points is sufficient to achieve good classification performance for GAN-synthesized faces.

Original languageEnglish
Title of host publicationIH and MMSec 2019 - Proceedings of the ACM Workshop on Information Hiding and Multimedia Security
PublisherAssociation for Computing Machinery, Inc
Pages113-118
Number of pages6
ISBN (Electronic)9781450368216
DOIs
StatePublished - Jul 2 2019
Event7th ACM Workshop on Information Hiding and Multimedia Security, IH and MMSec 2019 - Paris, France
Duration: Jul 3 2019Jul 5 2019

Publication series

NameIH and MMSec 2019 - Proceedings of the ACM Workshop on Information Hiding and Multimedia Security

Conference

Conference7th ACM Workshop on Information Hiding and Multimedia Security, IH and MMSec 2019
Country/TerritoryFrance
CityParis
Period07/3/1907/5/19

Keywords

  • Facial landmarks
  • GANs
  • Image forensics

Fingerprint

Dive into the research topics of 'Exposing GAN-synthesized faces using landmark locations'. Together they form a unique fingerprint.

Cite this