eduzhai > Applied Sciences > Engineering >

Facial Expression Editing with Continuous Emotion Labels

  • Save

... pages left unread,continue reading

Document pages: 8 pages

Abstract: Recently deep generative models have achieved impressive results in the fieldof automated facial expression editing. However, the approaches presented sofar presume a discrete representation of human emotions and are thereforelimited in the modelling of non-discrete emotional expressions. To overcomethis limitation, we explore how continuous emotion representations can be usedto control automated expression editing. We propose a deep generative modelthat can be used to manipulate facial expressions in facial images according tocontinuous two-dimensional emotion labels. One dimension represents anemotion s valence, the other represents its degree of arousal. We demonstratethe functionality of our model with a quantitative analysis using classifiernetworks as well as with a qualitative analysis.

Please select stars to rate!


0 comments Sign in to leave a comment.

    Data loading, please wait...