BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Australia/Melbourne
X-LIC-LOCATION:Australia/Melbourne
BEGIN:DAYLIGHT
TZOFFSETFROM:+1000
TZOFFSETTO:+1100
TZNAME:AEDT
DTSTART:19721003T020000
RRULE:FREQ=YEARLY;BYMONTH=4;BYDAY=1SU
END:DAYLIGHT
BEGIN:STANDARD
DTSTART:19721003T020000
TZOFFSETFROM:+1100
TZOFFSETTO:+1000
TZNAME:AEST
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20260114T163710Z
LOCATION:Meeting Room C4.9+C4.10\, Level 4 (Convention Centre)
DTSTART;TZID=Australia/Melbourne:20231213T123000
DTEND;TZID=Australia/Melbourne:20231213T124500
UID:siggraphasia_SIGGRAPH Asia 2023_sess163_papers_274@linklings.com
SUMMARY:Object Motion Guided Human Motion Synthesis
DESCRIPTION:Jiaman Li, Jiajun Wu, and Karen Liu (Stanford University)\n\nM
 odeling human behaviors in contextual environments has a wide range of app
 lications in character animation, embodied AI, VR/AR, and robotics. In rea
 l-world scenarios, humans frequently interact with the environment and man
 ipulate various objects to complete daily tasks. In this work, we study th
 e problem of full-body human motion synthesis for the manipulation of larg
 e-sized objects. We propose Object MOtion guided human MOtion synthesis (O
 MOMO), a conditional diffusion framework that can generate full-body manip
 ulation behaviors from only the object motion. Since naively applying diff
 usion models fails to precisely enforce contact constraints between the ha
 nds and the object, OMOMO learns two separate denoising processes to first
  predict hand positions from object motion and subsequently synthesize ful
 l-body poses based on the predicted hand positions. By employing the hand 
 positions as an intermediate representation between the two denoising proc
 esses, we can explicitly enforce contact constraints, resulting in more ph
 ysically plausible manipulation motions. With the learned model, we develo
 p a novel system that captures full-body human manipulation motions by sim
 ply attaching a smartphone to the object being manipulated. Through extens
 ive experiments, we demonstrate the effectiveness of our proposed pipeline
  and its ability to generalize to unseen objects. Additionally, as high-qu
 ality human-object interaction datasets are scarce, we collect a large-sca
 le dataset consisting of 3D object geometry, object motion, and human moti
 on. Our dataset contains human-object interaction motion for 15 objects, w
 ith a total duration of approximately 10 hours.\n\nRegistration Category: 
 Full Access\n\nSession Chair: Chek Tien Tan (Singapore Institute of Techno
 logy, Centre for Immersification)\n\n
URL:https://asia.siggraph.org/2023/full-program?id=papers_274&sess=sess163
END:VEVENT
END:VCALENDAR