Taein Kwon

Taein Kwon is a postdoctoral research fellow at the Visual Geometry Group (VGG) within the Department of Engineering Science at the University of Oxford, working under Prof. Andrew Zisserman. Previously, he completed his PhD at ETH Zurich under Prof. Marc Pollefeys and earned master's and bachelor's degrees from UCLA and Yonsei University, respectively.

His educational background includes:

• Bachelor's in Electrical Engineering from Yonsei University, Seoul, Korea
• Master's degree from UCLA
• PhD from ETH Zurich (defended July 2024)

His research spans Egocentric Vision, Action Recognition, Hand-object Interaction, Video Understanding, AR/VR, and Multi-modal Learning, with emphasis on first-person perspective analysis for AI assistants and human-computer interaction. His work integrates 3D reconstruction, pose estimation, and multimodal signals to model complex human activities and physical interactions.

Analysis of his 2021-2025 publications reveals a consistent focus on egocentric vision datasets (H2O, HoloAssist, EgoPressure) and novel frameworks for hand-object interaction, action recognition, and gesture understanding. His research demonstrates strong interdisciplinary connections between computer vision, robotics, and human-centered AI, with increasing emphasis on pressure sensing, co-speech gestures, and cross-modal alignment.

His scientific recognition includes:

• CVPR Egovis 2022/2023 Distinguished Paper Award for HoloAssist (July 2024)
• SNSF Postdoc.Mobility fellowship (May 2024)

He actively mentors students on egocentric vision projects, supervising master's theses, semester projects, and collaboration initiatives leading to publications at top conferences. His research is supported by the SNSF fellowship and industry collaborations with Meta Reality Labs and Microsoft Research.

As part of Oxford's Visual Geometry Group, he contributes to cutting-edge computer vision research while maintaining strong ties with ETH Zurich's computer vision community through ongoing collaborations and dataset development efforts.