ECCV 2018 Tutorial

Functional Maps

A Flexible Representation for Learning and Computing Correspondence

Course overview

This course will introduce the audience to the techniques for learning, computing and processing correspondences between shapes, understood in a broad sense as domain or signal geometry, proximity or connectivity (e.g. images, point clouds, meshes or graphs) based on the functional map representation. We will provide the mathematical background, computational methods and various applications of this framework in computer vision and machine learning problems.

Intended audience

This course is intended for students, researchers, and practitioners in computer vision / pattern recognition / machine learning dealing with problems related to map inference, information transport and correspondence between and across geometric datasets (e.g., flat images, low- and high-dimensional pointclouds, polygonal meshes, volumetric images, graphs).

When and where

Theresianum 1601, Technische Universität München Arcisstraße 21, 80333 Munich


  • 8:20 - 8:30 Introduction [10 min] (Emanuele Rodolà)
  • 8:30 - 9:30 Motivation + basic properties of the functional map representation [60 min] (Maks Ovsjanikov) (slides)
  • 9:30 - 10:30 Shape differences, functional map networks [60 min] (Leonidas Guibas) (slides)
  • 10:30 - 11:00 Coffee Break
  • 11:00 - 12:30 Partiality, clutter, and learning with functional maps [90 min] (Emanuele Rodolà) (slides)


Maks Ovsjanikov, Ecole Polytechnique

Emanuele Rodolà, Sapienza University of Rome

Or Litany, Stanford / Facebook AI Research

Leo Guibas, Stanford