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:20240214T070250Z
LOCATION:Meeting Room C4.11\, Level 4 (Convention Centre)
DTSTART;TZID=Australia/Melbourne:20231215T162000
DTEND;TZID=Australia/Melbourne:20231215T163500
UID:siggraphasia_SIGGRAPH Asia 2023_sess139_papers_313@linklings.com
SUMMARY:Reconstructing Close Human Interaction from Multiple Views
DESCRIPTION:Technical Papers\n\nQing Shuai (Zhejiang University); Zhiyuan 
 Yu (Department of Mathematics, Hong Kong University of Science and Technol
 ogy); Zhize Zhou (Capital University of Physical Education and Sports); Li
 xin Fan and Haijun Yang (WeBank); Can Yang (Department of Mathematics, Hon
 g Kong University of Science and Technology); and Xiaowei Zhou (State Key 
 Laboratory of CAD&CG, Zhejiang Univerisity)\n\nThis paper addresses the ch
 allenging task of reconstructing the poses of multiple individuals engaged
  in close interactions, captured by multiple calibrated cameras. The diffi
 culty arises from the noisy or false 2D keypoint detection due to inter-pe
 rson occlusion, the heavy ambiguity to associate keypoints to individuals 
 due to the close interactions, and the scarcity of training data, as colle
 cting and annotating extensive data in crowd scenes is resource-intensive.
  \n\nWe introduce a novel learning-based system to address these challenge
 s. Our approach constructs a 3D volume from multi-view 2D keypoint heatmap
 s, which is then fed into a conditional volumetric network to estimate the
  3D pose for each individual.\n\nAs the network doesn't need images as inp
 ut, we can leverage known camera parameters from test scenes and a large q
 uantity of existing motion capture data to synthesize massive training dat
 a that mimics the distribution of the real data in the test scenes.\n\nExt
 ensive experiments across various camera setups and population sizes demon
 strate that our approach significantly surpasses previous approaches in te
 rms of both pose accuracy and generalizability. The code will be made publ
 icly available upon acceptance of the paper.\n\nRegistration Category: Ful
 l Access\n\nSession Chair: Yuting Ye (Reality Labs Research, Meta)
URL:https://asia.siggraph.org/2023/full-program?id=papers_313&sess=sess139
END:VEVENT
END:VCALENDAR