From Vision to Language and Back Again
Andrei Barbu, Research Scientist, Massachusetts Institute of Technology (MIT)
Wednesday, November 15, 1:00-2:00pm; Hariri Institute for Computing (111 Cummington Mall, Boston)
Abstract: We will discuss a program to unify research around a number of vision-language problems into a single mathematical framework culminating in a robotic platform that is able to follow natural language commands, store knowledge, and answer questions. Connecting perception and language in diverse real-world settings sheds light on how cognition is structured and can help answer questions about the nature of memory. We will look at near-term extensions of this work to allow robots to introspect, detect failure, and take corrective action. I will also briefly discuss the state of science in computer vision and present a new soon-to-be-released object-detection dataset and new methodology for evaluating computer vision systems.
Bio: Andrei Barbu is a Research Scientist at MIT in the Computer Science and Artificial Intelligence Laboratory (CSAIL) working on vision, language, robotics, and neuroscience. Dr. Barbu’s research interests focus around how flexible intelligence arises from the integration of perception, communication, and reasoning. Dr. Barbu received his his undergraduate degree from University of Waterloo and PhD from Purdue University.