HeadGAN: A Neural Network for One-Shot Reenactment

Check out HeadGAN – a new neural network that performs one-shot reenactment by fully transferring the facial expressions and head pose from a driving frame to a reference image.

Michail Christos Doukas, Stefanos Zafeiriou, and Viktoriia Sharmanska have introduced HeadGAN – a new neural network that performs one-shot reenactment by fully transferring the facial expressions and head pose from a driving frame to a reference image. It can be used for facial video compression and reconstruction. In addition, HeadGAN can be applied to facial expression editing, novel view synthesis, and face frontalization.

"We propose a GAN-based system that conditions synthesis on 3D face representations, which can be extracted from a driving video and adapted to the facial geometry of the reference image," comments the team. "We capitalize on prior knowledge of expression and identity disentanglement, enclosed within 3D Morphable Models (3DMMs)."

You can learn more about HeadGAN here. Also, don't forget to join our new Reddit page, our new Telegram channel, follow us on Instagram and Twitter, where we are sharing breakdowns, the latest news, awesome artworks, and more.

HeadGAN: A Neural Network for One-Shot Reenactment

Join discussion

Comments 0

You might also like

We need your consent