Discover how machine learning can be used to model the way that faces move during speech. In this talk, UEA’s Dr Sarah Taylor will explain how these techniques enable us to automatically animate both computer generated characters and the faces of real humans in real video. You will learn that seeing is not always believing!
Practical applications of speech animation – computer games and animated films, for example – often rely on motion capture devices or laborious hand animation. Both of these approaches are expensive and time consuming. These challenges have motivated the Graphics, Vision and Speech research group at UEA to investigate automatic approaches to speech animation that would both speed up, and improve the quality of existing methods.
In this talk, you will learn about the complexity of human faces during speech and why automatic speech animation is more challenging than it might initially seem. You’ll discover how machine learning techniques can be used to train computational models to learn how talking faces move, and subsequently how these models can control the faces of 3D characters. Finally, you will see how the same approach can be used to animate not just cartoon faces, but the faces of real people in real video.
Dr Sarah Taylor is a Lecturer in Computing Sciences in the School of Computing Sciences at the University of East Anglia.
Friday 25 October
Venue: The Forum, Auditorium
Cost: Free, booking required
Tickets will go off sale at 7pm the night before the event. For events at The Forum and Sir Isaac Newton, any remaining tickets will be available for sale on the day of the event at our ticket desk in the Atrium. For events at other venues, remaining tickets will be available to purchase, cash only, on the door 30 mins before the start time.
One ticket required per person.