Presentation

Diffeomorphic Neural Surface Parameterization for 3D and Reflectance Recovery
SessionRoundtable Session: Neural Objects, Materials and Illumination
Authors
Ziang Cheng
Hongdong Li
Richard Hartley
Yinqiang Zheng
Imari Sato
Event Type
Technical Paper
Keywords
Rendering
Interest Areas
Research & Education
Presentation Types
In Person
Registration Categories
Full Conference Supporter
Full Conference
Exhibitor Additional Full Conference
Exhibitor Full Conference
Recordings
This session WILL NOT be recorded.
TimeMonday, 8 August 202211:01am - 11:06am PDT
LocationEast Building, Ballroom A/B
DescriptionWe present a neural approach for inverse rendering. Key to our method is a velocity field surface parameterization technique that is provably diffeomorphic and orientation preserving. Such surface representation is coded in a recurrent residual network structure and allows a surface reflectance map to be simultaneously recovered to support novel rendering.
Authors
Ziang Cheng
The Australian National University
Hongdong Li
The Australian National University
Richard Hartley
The Australian National University
Yinqiang Zheng
The University of Tokyo
Imari Sato
National Institute of Informatics