Deep Learning and Geometry

Outline

The past decade in computer vision research has witnessed the re-emergence of “deep learning”, and in particular convolutional neural network (CNN) techniques, allowing to learn powerful image feature representations from large collections of examples. Such methods have achieved a breakthrough in performance in a wide range of applications such as image classification, segmentation, detection and annotation. Nevertheless, when attempting to apply standard deep learning methods to geometric data which by its nature is non-Euclidean (e.g. 3D shapes and graphs), one has to face fundamental differences between images and geometric objects. Shape analysis, graph analysis, and geometry pose new challenges that are non-existent in image analysis, and deep learning methods have only recently started penetrating into the 3D vision, pattern recognition, multimedia, signal processing, and graphics communities. Deep learning has been applied to 3D data in recent works using standard (Euclidean) architectures applied to volumetric or view-based shape representations. Intrinsic versions of deep learning have also been proposed very recently with the generalization of the CNN paradigm to non-Euclidean manifolds, allowing them to deal with domain deformations. These “generalized” CNNs can be used to learn invariant shape features and correspondence, allowing to achieve state-of-the-art performance in several shape analysis tasks, while at the same time allowing for different shape representations, e.g. meshes, point clouds, or graphs.

The main focus of the workshop is on generalization of deep learning techniques beyond the Euclidean settings, in order to apply them to geometric data. We aim to bring together and offer a forum for discussion and interaction among researchers interested in learning techniques applied to geometric data, and favor the cross-fertilization between fields such as Machine Learning, Computer Graphics, Computer Vision, Signal Processing, Pattern Recognition, and Multimedia. We believe that the workshop will give a new view on deep learning and thus will be interesting to machine learning experts on the one hand, and will offer new solutions to hard problems in signal processing, pattern recognition, and computer graphics and thus would appeal to experts in these fields on the other.


Organizers

Main contact: Or Litany (Tel Aviv University) <orlitany at gmail dot com >

Co-organizers: Emanuele Rodolà (USI Lugano) , Michael Bronstein (USI Lugano), Alex Bronstein (Technion), Ron Kimmel (Technion)


Important dates

Full paper submission: June 8, 2017 June 19, 2017

Notification of acceptance: July 4, 2017

Camera-ready paper: July 20, 2017

Workshop (half day): September 2, 2017


Submission

Please follow the submission guidelines