Within this project, the goal is to evaluate two concurrent approaches for the animation of realistic facial motion using (a) a conventional linear face model and (b) a novel non-linear face model based on fundamental coordinates. Based on a large data set, (non)-linear motion trajectories will be estimated within the respective model spaces. Both approaches are to be evaluated quantitatively and with respect to their visual difference.

Please look at the announcement for more details.

Geometric Deep Learning for Dense Correspondence