skip to main content

H.B. Keller Colloquium

Monday, October 23, 2023
4:00pm to 5:00pm
Add to Cal
Annenberg 105
Learning to Perceive in the Era of Large Models
Georgia Gkioxari, Assistant Professor of Computing and Mathematical Sciences and Electrical Engineering, CMS - EE, Caltech,

I can't think of a more exciting time to be a computer scientist! The rise of data-driven computational methods hasn't just shown potential; they've spearheaded remarkable advancements in the world of computing. For the lighthearted, this means any one of us can now dub as a painter or a poet (DaVinci would maybe disagree!)! On a more profound level, harnessing data has enabled us to address previously unthinkable challenges. Perception, the task of translating images into meaningful concepts, is a prime example. Perception entails converting a grid of red-green-blue pixels to objects, their semantics, their interactions and their 3D structure – this is not a simple feat. In this talk, I'll walk you through my endeavors in visual perception, from identifying and localizing objects to predicting their 3D shape and position, all from just one image. To achieve this, I build models that learn from data. But data does not always come in the form we desire and can often pose significant computational challenges. My work develops innovative models and tools to bridge the gap between the digital two-dimensional realm to the continuous 3D world while discerning objects, their semantics and geometry. These advancements have been remarkable, but the future of data-centric computing looks even brighter thanks to the emerging tools and ever-evolving ways to learn from data, be it scarce or abundant.

For more information, please contact Sumaia Abedin by phone at 6263956704 or by email at [email protected] or visit