Events

Efficient learning on curved surfaces via diffusion

Speaker: Maks Ovsjanikov

Location: 60 Fifth Avenue, Room 527

Date: Thursday, June 30, 2022

In this talk I will describe several approaches for learning on curved surfaces, represented as point clouds or triangle meshes. I will first give a brief overview of geodesic convolutional neural networks (GCNNs) and then present a recent approach that replaces this paradigm with a framework based on learned diffusion. The key properties of this approach are its efficiency, intrinsic nature and robustness to changes in discretization. I will then show several applications, ranging from RNA surface segmentation to non-rigid shape matching. Finally, I will describe a recent learning-based method that uses this architecture, while addressing partial non-rigid shape correspondence.