Presentation
DiffusionNet: Discretization Agnostic Learning on Surfaces
SessionTechnical Papers
Presenter
Event Type
Technical Paper
Research & Education
Virtual
Full Conference Supporter
Full Conference
Virtual Conference Supporter
Virtual Conference
Exhibitor Additional Full Conference
Exhibitor Full Conference
Time
Location
DescriptionDiffusionNet is a general-purpose method for deep learning on meshes, points clouds, etc., via simple spatial diffusion. The networks generalize by construction across different samplings, resolutions, and representations of shapes and scale to high-resolution inputs. Spatial support is automatically optimized as a continuous parameter.
Presenter
