The animation of digital character faces in movies, TV productions and social interactions requires complex hardware setups (head-mounted camera) or a high number of hours of manual work by artists through the manipulation of so-called blendshapes.
The purpose of this internship is to investigate an alternative, where the facial animation would be generated using a deep generative network (GAN) drive by text or audio signals (i.e. text or speech to facial animation).
A first objective will be to evaluate performances and limitations of existing solutions, and try one (for instance temporal GAN). Then depending on the results, either develop a new deep architecture to compensate the current limitations, or ingest some emotional features into a network for emotional driven facial animation (such as in https://arxiv.org/pdf/1908.03904.pdf).
The Internship is located at InterDigital Research & Innovation, Rennes, france.