Keynotes
Shalini De Mello
Distinguished Research Scientist - Nvidia Research
Title
Do We Need 3D Data to Learn Geometry?
Abstract
Two-dimensional (2D) generative models based on GANs and, more recently, diffusion models have revolutionized photorealistic image and video synthesis. These large models, trained with Internet-scale data, have the ability to compose multiple different concepts together to generate high-quality and semantically plausible images and videos of complex scenes. These advances have also sparked much interest recently in synthesizing 3D content using 2D generative models or images alone by coupling them with neural volumetric representations. These novel approaches beg the question, “Do we need 3D data to learn geometry?” To answer this question, in this talk, we will explore how far methods that rely only on 2D images have come in enabling 3D content synthesis. We will examine their primary mechanisms including, 3D-aware GANS and score distillation sampling with diffusion guidance; summarize the current performance of the state of the art and its shortcomings; and provide real-world examples of how such 3D synthesized data can enable new downstream applications for geometric processing.
Biography
Dr. Shalini De Mello is a Distinguished Research Scientist in the Learning and Perception Research group at NVIDIA, which she joined in 2013. Her research interests are in human-centric vision (face and gaze analysis) and in data-efficient (synth2real, low-shot, self-supervised and multimodal) machine learning. She has co-authored 55 peer-reviewed publications and 44 patents. Her inventions have contributed to several NVIDIA products, including DriveIX and Maxine. Previously, she has worked at Texas Instruments and AT&T Laboratories. She received her Doctoral degree in Electrical and Computer Engineering from the University of Texas at Austin.
Olga Sorkine-Hornung
Professor of Computer Science, ETH Zurich
Title
20 Years of mesh editing: what we learned and what to learn.
Biography
Dr. Olga Sorkine-Hornung is a Professor of Computer Science at ETH Zurich, Switzerland, where she leads the Interactive Geometry Lab at the Institute of Visual Computing. Prior to joining ETH, she was an Assistant Professor at the Courant Institute of Mathematical Sciences, New York University (2008-2011). She earned her BSc in Mathematics and Computer Science and PhD in Computer Science from Tel Aviv University (2000, 2006). Following her studies, she received the Alexander von Humboldt Foundation Fellowship and spent two years as a postdoc at the Technical University of Berlin. Her primary research interests lie in computer graphics, geometry processing, shape representation and modelling and discrete differential geometry. She has published over 140 peer-reviewed articles in top computer science venues. Olga has served as a member of the ACM Turing Award Committee, including serving as the ACM Turing Award Committee Chair for 2020. She received several scientific prizes and awards, including the ERC Starting and Consolidator grants, the ACM SIGGRAPH Significant New Researcher award and the Eurographics Outstanding Technical Contributions award. She is a Fellow of the ACM and the Eurographics Association and a member of the Swiss Academy of Engineering Sciences (SATW).
Julie Digne
Senior Researcher, CNRS, Lyon, France
Title
Machine Learning for Geometric Shape Analysis
Abstract
Neural Implicits have recently attracted a lot of attention as a way to represent shapes and avoid any explicit representation. In this talk I will describe how to leverage this description for shape analysis tasks and in particular global information such as medial axis extraction. In a second part of the talk, I will discuss how to leverage geometric information to improved image segmentation tasks yielding lighter and faster to train segmentation networks.
Biography
Dr. Julie Digne is a Senior Researcher at CNRS. Her main interests are Geometry Processing, and more precisely surface analysis, denoising, compression, and segmentation, using conventional or machine learning based approaches. She obtained her PhD in Applied Mathematics from École Normale Supérieure Paris-Saclay (France) in 2010 with Jean-Michel Morel, for which she was awarded the Fondation Jacques Hadamard PhD award. She publishes in top venues for Computer Graphics, Geometry, and Machine Learning, and serves in top conference committees (Siggraph, Eurographics, SGP). In 2021 she was program chair of the Symposium on Geometry Processing.