일시: 2026. 7. 27. (월요일), 오후 1시~2시

장소: 제4공학관 D403호

Speaker: Youngjoong Kwon (권영중), Ph.D. / Assistant Professor, Department of Computer Science, Emory University

Title: Everyday Human Digitization: The Bridge to 4D World Model

Abstract:
Recent advances toward 4D world models have produced remarkably realistic videos. However, visual realism does not necessarily imply an understanding of how people interact with the world. Building useful 4D world models therefore requires accurate modeling of humans outside controlled studio environments.

This talk focuses on everyday human digitization through three challenges: efficiency, affordability, and scalability. For efficiency, I present a deformable light-field representation that supports real-time rendering of animatable humans. For affordability, I introduce a monocular performance-capture method that progressively accumulates temporal observations in a canonical memory to reconstruct regions that are not visible in the current frame. For scalability, I discuss how in-the-wild videos can improve generalization to the real-world data distribution, including diverse body shapes, appearances, and clothing. Together, these works move human digitization beyond specialized capture studios and toward practical capture with ordinary cameras.

Bio:
Youngjoong Kwon is an Assistant Professor of Computer Science at Emory University. Her research focuses on 3D and 4D human digitization, human performance capture, neural rendering, and generative models. Before joining Emory, she was a postdoctoral scholar at Stanford University and Carnegie Mellon University. She received her Ph.D. in Computer Science from the University of North Carolina at Chapel Hill and her B.S. in Computer Science from Yonsei University.