BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Taming foundation models for visual concept learning and 4D modeli
 ng - Kai Han\, University of Hong Kong
DTSTART:20250911T100000Z
DTEND:20250911T110000Z
UID:TALK235939@talks.cam.ac.uk
CONTACT:Elliott Wu
DESCRIPTION:In this talk\, I will present our recent work on leveraging fo
 undation models for open-world visual concept learning and 4D modeling. Fi
 rst\, I will discuss how we repurpose vision foundation models for continu
 al category discovery by learning a flexible Gaussian mixture prompt pool.
  Next\, I will introduce our approach to automatically extracting visual c
 oncepts\, both at the object and intrinsic levels\, using Stable Diffusion
  models. Finally\, I will share our work on high-quality 4D generation by 
 effectively harnessing video diffusion models\, enabling temporally and sp
 atially consistent content creation with 4D Gaussian splatting.
LOCATION: Cambridge University Engineering Department\, JDB Teaching Room
END:VEVENT
END:VCALENDAR