Abstract

Abstract

Understanding the 3D structure of real-world environments is a fundamental challenge in machine perception, with many applications towards robotic navigation and interaction, content creation, and mixed reality scenarios. In this talk, we propose to develop an object-centric perspective on 3D perception. We first explore 3D CAD -based priors to represent objects to attain a compact representation space from which we can develop robust retrieval and alignment to image observations. We then relax the constraint on retrieval from a database to learn parametric models to represent shapes and parts, followed by developing a parametric model for the manifold of 3D scenes. Finally, we show that an object basis not only helps to perform semantic reconstruction, but can also significantly benefit camera tracking for RGB -D video observations—providing the potential for holistic reconstruction and understanding from an object-guided perspective.

Bio

Angela Dai is an Assistant Professor at the Technical University of Munich where she leads the 3D AI group. Prof. Dai’s research focuses on understanding how the 3D world around us can be modeled and semantically understood. Previously, she received her PhD in computer science from Stanford in 2018 and her BSE in computer science from Princeton in 2013. Her research has been recognized through a Eurographics Young Researcher Award, Google Research Scholar Award, ZDB Junior Research Group Award, an ACM SIGGRAPH Outstanding Doctoral Dissertation Honorable Mention, as well as a Stanford Graduate Fellowship.

Location

The talk will be given at LR3B in the Engineering Department (Trumpington St, Cambridge CB2 1PZ ).

Google Calendar

To get updates on future seminars, please subscribe to the following Google calendar: https://calendar.google.com/calendar/u/0?cid=c2pjcHN0YXM2N3QyMWU3c2FqNjBqNWNiYXNAZ3JvdXAuY2FsZW5kYXIuZ29vZ2xlLmNvbQ