Centre for Intelligent Sensing seminar

Title of talk: 

Egocentric Understanding: A Path to Context-Aware AI

Egocentric understanding plays a crucial role in enabling context-aware AI, especially in Mixed Reality environments and AI robotics, where perception is grounded in an egocentric view. Despite its importance, the field faces two key challenges: a shortage of high-quality, multimodal egocentric video datasets and the limited capability of current systems to effectively process these complex data streams. In this talk, I’ll walk through the full pipeline of egocentric understanding, from dataset collection to the development of algorithms that can effectively interpret egocentric videos to tackle these challenges and pave the way for context-aware AI.

Taein Kwon is a postdoctoral fellow at the Visual Geometry Group (VGG), University of Oxford, working with Prof. Andrew Zisserman. His research interests include egocentric understanding, hand-object interaction, video understanding, multimodal human-AI interaction, AR/VR/MR, robotics, and computer vision. He has published several papers on these topics at leading computer vision conferences such as ICCV, CVPR, and ECCV. He also co-organized a workshop on Human Body, Hands, and Activities from Egocentric and Multi-view Cameras at ECCV 2022 and served as the HoloAssist Challenge Chair at the EgoVis Workshop, CVPR 2024. He earned a Ph.D. in Computer Science from ETH Zurich, an M.S. in Electrical and Computer Engineering from UCLA, and a B.S. in Electrical Engineering from Yonsei University in Seoul, Korea.