Keynote Speakers

Title: Generation, Animation, and Evaluation of Realistic Virtual Humans

Abstract: Realistic virtual humans have many interesting applications in several fields, ranging from characters in computer games, to avatars representing ourselves in virtual environments, and digital models for medical applications. In this talk I will first present our avatar generation pipeline, which produces a realistic ready-to-animate avatar of a real person in just a couple of 10 minutes. I will show how the resulting avatars improve the perceived presence and virtual body ownership in virtual reality experiments, thereby confirming the need for realistic avatars. Finally, I will present a physics-based skinning technique that enables realistic character animation with dynamic effects while still achieving real-time frame rates.

Title: Functional Approaches for Representing and Analyzing 3D Shapes With and Without Learning

Abstract: I will give an overview of a set of techniques for representing and analyzing 3D shapes via spaces of real-valued functions defined on them. While this point of view is classical in many areas of mathematics, it has only recently been applied in geometric shape modeling. I will give an overview of these and related techniques and demonstrate their utility in problems including shape matching and shape comparison. A key insight is that both shapes and relations between them can be represented and manipulated as small-sized matrices, which naturally fits within many modern learning pipelines. Finally, I will show very recent deep-learning based data-driven methods for both shape correspondence and shape synthesis that exploit this functional representation to lead to state-of-the-art results.