As a camera moves through its environment it is possible to build and map and localise the camera with respect to the map in real-time, as process known as Visual SLAM. The geometry of this is well understood, but cameras have the ability to deliver much more than geometry, especially through using learning to understand scene semantics and using data-driven prior knowledge. In this talk I will discuss ideas and progress in leveraging the power of deep learning to go beyond geometric SLAM; to leverage prior domain knowledge to improve map and shape estimation; to enable navigation in previously unvisited environments; and to use deep-learned scene representations. The ultimate goal is to turn the camera into a device for flexible, large-scale situational awareness -- or as some have called it, a "Spatial AI" sensor.

About the speaker:
Ian Reid is a Professor of Computer Science at the University of Adelaide, and he was Head of School 2018-2022. He is a Fellow of the Australian Academy of Technological Sciences and Engineering (ATSE), Fellow of the Australian Academy of Science (AAS), former Rhodes Scholar, and held an Australian Laureate Fellowship 2013-2018. He was the Deputy Director of the Australian Centre for Robotic Vision 2014-2021.