GTC 2020: Generative Face Models from Light Stage Scans
After clicking “Watch Now” you will be prompted to login or join.
Generative Face Models from Light Stage Scans
Kalle Bladin, USC Institute for Creative Technologies | Yajie Zhao, USC Institute For Creative Technologies
GTC 2020
We'll discuss how GPUs enable research based on facial-scan data from the USC Institute for Creative Technologies' Light Stage X for virtual avatar creation. ICT's Vision and Graphics Lab has devised a new database of facial-scan data that lends itself to machine learning-based approaches for inferring relightable face models of automatically rigged virtual avatars using a non-linear morphable model. Additionally, new advances in deployable, fast, facial-tracking solutions make interactive face-to-face communication simpler. We'll describe some of our success using NVIDIA's new face-tracker SDK, and also touch on some advances in real-time rendering of human faces.