22nd EANN 2021, 25 - 27 June 2021, Greece

Automatic Facial Expression Neutralisation Using Generative Adversarial Network

Wiem Grina, Ali Douik


  Face recognition systems has reached a level of maturity which makes it applicable in many applications. However, despite these recent results, the precision of face recognition systems still needs to be improved, especially with regard to facial expression and pose variation. Unlike existing methods that mostly operate on existed databases for recognition and identification, we present a new technique of facial expression neutralisation based on synthesizing realistic neutral images using Generative Adversarial Networks (GANs).In this article, we want to propose a new optimization technique called Ranger to be applied in the GANnotation architecture. Promising results are obtained on the RaFD dataset leading to improvements in the facial recognition system.  

*** Title, author list and abstract as seen in the Camera-Ready version of the paper that was provided to Conference Committee. Small changes that may have occurred during processing by Springer may not appear in this window.