Presentation / Installation
StyleAvatar: Real-time Photo-realistic Neural Portrait Avatar From a Single Video
DescriptionWe propose StyleAvatar, a real-time, photo-realistic portrait avatar reconstruction method using StyleGAN-based networks. We introduce a compositional representation and a sliding window augmentation method to enable faster convergence and improve translation generalization. The proposed networks can converge within two hours with high image quality and support building a real-time system.
Event Type
Technical Paper
TimeThursday, 10 August 202311:06am - 11:16am PDT
LocationPetree Hall D
ACM Digital Library
Technical Paper PDF
Session Time & Location
Research & Education
Livestreamed
Recorded
Animation/Simulation
Modeling
Video
FC
FCS
V
VS
EFC