Passer au contenu

/ Département d'informatique et de recherche opérationnelle

Je donne

Rechercher

Navigation secondaire

Neural Shape Mapping: Modifying, Manipulating, and Matching Geometry Through Deep Learning

Neural Shape Mapping: Modifying, Manipulating, and Matching Geometry Through Deep Learning

Par

Noam Aigerman

Research Scientist at Adobe Research

 

vendredi 24 février 2023, 10:30-12:00 ESTSalle 3195


Pavillon André-Aisenstadt, Université de Montréal, 2920 Chemin de la Tour

 

Abstract: In this talk I will discuss my recent efforts in developing techniques for AI-driven shape mapping, i.e., enabling AI frameworks to manipulate, modify, and retarget 3D geometry. This task has the potential to automate many critical applications that involve 3D shapes, such as 3D design, medical imaging, video-to-3D, and physical simulation, to name a few. However, achieving AI-driven mapping capabilities is a highly-intricate problem, due to the inherent difficulties in applying deep-learning techniques to 3D geometry. This difficulty has been, to a significant degree, overcome by one of my latest works, which provides a fundamental, versatile framework for AI-driven mapping of shapes through neural jacobian fields. This will be the focal point of my talk, along with a complementary work which tackles the more specific (and restrictive) setting of computing surface-to-surface mappings.


Bio: Noam Aigerman is a Research Scientist at Adobe Research, focusing on fundamental research at the intersection of geometry processing, numerical optimization, and machine learning. He joined Adobe after completing his PhD at the Weizmann Institute of Science, under the supervision of Prof. Yaron Lipman.