BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Asia/Hong_Kong
X-LIC-LOCATION:Asia/Hong_Kong
BEGIN:STANDARD
TZOFFSETFROM:+0800
TZOFFSETTO:+0800
TZNAME:HKT
DTSTART:19911015T033000
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20251218T030653Z
LOCATION:Meeting Room S221\, Level 2
DTSTART;TZID=Asia/Hong_Kong:20251216T151100
DTEND;TZID=Asia/Hong_Kong:20251216T152200
UID:siggraphasia_SIGGRAPH Asia 2025_sess127_papers_2309@linklings.com
SUMMARY:Gaussian See, Gaussian Do: Semantic 3D Motion Transfer from Multiv
 iew Video
DESCRIPTION:Yarin Bekor (Technion - Israel Institute of Technology); Gal M
 ichael Harari (Technion – Israel Institute of Technology); Or Perel (NVIDI
 A, University of Toronto); and Or Litany (Technion – Israel Institute of T
 echnology, NVIDIA)\n\nWe present Gaussian See, Gaussian Do, a novel approa
 ch for semantic 3D motion transfer from multiview video. Our method enable
 s rig-free, cross-category motion transfer between objects with semantical
 ly meaningful correspondence. Building on implicit motion transfer techniq
 ues, we extract motion embeddings from source videos via condition inversi
 on, apply them to rendered frames of static target shapes, and use the res
 ulting videos to supervise dynamic 3D Gaussian Splatting reconstruction. O
 ur approach introduces an anchor-based view-aware motion embedding mechani
 sm, ensuring cross-view consistency and accelerating convergence, along wi
 th a robust 4D reconstruction pipeline that consolidates noisy supervision
  videos. We establish the first benchmark for semantic 3D motion transfer 
 and demonstrate superior motion fidelity and structural consistency compar
 ed to adapted baselines.\n\nRegistration Category: Full Access, Full Acces
 s Supporter\n\nSession Chair: Tuur Stuyck (Meta)\n\n
END:VEVENT
END:VCALENDAR