Presentation

<
InvertAvatar: Incremental GAN Inversion for Generalized Head Avatars
Session3D Head Avatars From Data
DescriptionOur "Incremental 3D GAN Inversion" framework advances avatar reconstruction with flexible-number multiple frames, incorporating a unique animatable GAN prior for precise expressions control, UV parameterization for high-fidelity appearance recovery, and a ConvGRU-based recurrent network for temporal aggregation. We achieve photorealistic one-/few-shot 3D facial avatar modeling under one second.
Authors
Xiaochen Zhao
Tsinghua University
Jingxiang Sun
Tsinghua University
Lizhen Wang
Tsinghua University
Jinli Suo
Tsinghua University
Yebin Liu
Tsinghua University
Event Type
Technical Paper
TimeWednesday, 31 July 20249:20am - 9:30am MDT
LocationMile High 1
ACM Digital Library Journal Papers' PDFs
Conference Papers' PDFs
Session Time & Location
Sunday, 28 July 20246:00pm - 8:45pm MDTBluebird Ballroom
Wednesday, 31 July 20249:00am - 10:30am MDTMile High 1
Interest Areas
Research & Education
Recordings
Livestreamed
Recorded
Keywords
Animation
Geometry
Modeling
Rendering
Registration Categories
Full Conference
Full Conference Supporter
Virtual Access
Exhibitor Full Conference
Wednesday
Next PresentationNext Presentation3D Gaussian Blendshapes for Head Avatar Animation