Emanuele Rodolà
Emanuele Rodolà is Full Professor in Computer Science at Sapienza University of Rome, where he leads the GLADIA group on geometry, learning & applied AI. His research is funded by the ERC Starting Grant 2018 SPECGEO and a Google Research Scholar award.
Previously, he was a post-doctoral researcher at USI Lugano (2016-2017), an Alexander von Humboldt Fellow at TU Munich (2013-2016), and a JSPS Research Fellow at The University of Tokyo (2013). He received his PhD in Computer Science at Università Ca’ Foscari Venezia (2012), and graduated in Computer System Engineering at the University of Rome “Tor Vergata” (2008). During his doctoral studies, Dr. Rodolà spent a visiting research period at Tel Aviv University under the supervision of Prof. Alex Bronstein. He received a number of awards, including the Best Student Paper Award at 3DPVT 2010, the Best Paper Award at VMV 2015, and the Best Paper Award at SGP 2016. He has been serving in the program committees of the top rated conferences in computer vision (CVPR, ICCV, ECCV, ACCV, etc.), served as Area Chair at 3DV (2016, 2017), founded and chaired several workshops including the workshop on Geometry Meets Deep Learning (ECCV GMDL 2016, ICCV GMDL 2017), organized multiple SHREC contests, and was recognized (eight times) as IEEE Outstanding Reviewer, at CVPR (2013, 2015, 2016, 2017), ICCV (2015, 2017), and ECCV (2014, 2016). He gave tutorials and short courses in multiple occasions at EUROGRAPHICS, ECCV, SGP, SIGGRAPH, and SIGGRAPH Asia. His work on 3D reconstruction was featured by the national Italian television (RAI – Cose dell’altro Geo) in 2012.
Dr. Rodolà research interests include 3D shape analysis, matching, reconstruction and modeling.
Title: “From sound to metric priors: A new paradigm for shape generation”
Abstract: Spectral and metric geometry are at the heart of various problems in computer vision, graphics, pattern recognition, and machine learning. Ultimately, the core reason for their success can be traced down to questions of stability and to the informativeness of the eigenvalues of certain operators. In this talk, I will discuss and show tangible examples of such properties and showcase some dramatic implications on a selection of notoriously hard problems in computer vision and graphics. First, I will address the question of whether one can recover the shape of a geometric object from its vibration frequencies (‘hear the shape of the drum’); while theoretically the answer to this question is negative, little is known about the practical possibility of using the spectrum for shape reconstruction and optimization. I will introduce a numerical procedure called isospectralization, as well as a data-driven variant, showing how this *practical* problem is solvable. Then, I will discuss the increasingly popular task of designing an effective generative model for deformable 3D shapes. I will demonstrate how injecting metric distortion priors into a simple geometric reconstruction loss can lead to the formation of a very informative latent space, which can be trained with extremely scarce data (less than 10 examples) and still yield competitive generation quality as well as aiding geometric disentanglement.