Tensor-based Emotion Editing in the StyleGAN Latent Space

Research output: Contribution to conference - NOT published in proceeding or journalPaperResearchpeer-review


In this paper, we use a tensor model based on the Higher-Order Singular Value Decomposition (HOSVD) to discover semantic directions in Generative Adversarial Networks. This is achieved by first embedding a structured facial expression database into the latent space using the e4e encoder. Specifically, we discover directions in latent space corresponding to the six prototypical emotions: anger, disgust, fear, happiness, sadness, and surprise, as well as a direction for yaw rotation. These latent space directions are employed to change the expression or yaw rotation of real face images. We compare our found directions to similar directions found by two other methods. The results show that the visual quality of the resultant edits are on par with State-of-the-Art. It can also be concluded that the tensor-based model is well suited for emotion and yaw editing, i.e., that the emotion or yaw rotation of a novel face image can be robustly changed without a significant effect on identity or other attributes in the images.
Original languageEnglish
Publication date13 May 2022
Number of pages10
Publication statusPublished - 13 May 2022
EventAI for Content Creation Workshop
@ CVPR 2022
- New Orleans Ernest N. Morial Convention Center, New Orleans, United States
Duration: 19 Jun 202220 Jun 2022


ConferenceAI for Content Creation Workshop
@ CVPR 2022
LocationNew Orleans Ernest N. Morial Convention Center
Country/TerritoryUnited States
CityNew Orleans
Internet address


  • Computer vision and pattern recognition
  • Face Synthesis
  • generative adversarial network


Dive into the research topics of 'Tensor-based Emotion Editing in the StyleGAN Latent Space'. Together they form a unique fingerprint.

Cite this