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:20260114T163650Z
LOCATION:Meeting Room C4.8\, Level 4 (Convention Centre)
DTSTART;TZID=Australia/Melbourne:20231214T163500
DTEND;TZID=Australia/Melbourne:20231214T165000
UID:siggraphasia_SIGGRAPH Asia 2023_sess166_papers_629@linklings.com
SUMMARY:From Skin to Skeleton : Towards Biomechanically Accurate 3D Digita
 l Humans
DESCRIPTION:Marilyn Keller (Max Planck Institute for Intelligent Systems),
  Keenon Werling (Stanford University), Soyong Shin (Max-Planck-Institut fü
 r Informatik), Scott Delp (Stanford), Sergi Pujades (INRIA), Karen Liu (St
 anford University), and Michael Black (Max Planck Institute for Intelligen
 t Systems)\n\nGreat progress has been made in estimating 3D human pose and
  shape from images and video by training neural networks to directly regre
 ss the parameters of parametric human models like SMPL.\nHowever, existing
  body models have simplified kinematic structures that do not correspond t
 o accurate joint locations and articulations in the human skeletal system,
  limiting their potential use in biomechanics. On the other hand, methods 
 for estimating biomechanically accurate skeletal motion typically rely on 
 complex motion capture systems and expensive optimization methods.\nWhat i
 s needed is a parametric 3D human model with a biomechanically accurate sk
 eletal structure that can be easily regressed from images.\nTo that end, w
 e develop SKEL,  which re-rigs the SMPL body model with a biomechanics ske
 leton. To enable this, we need training data of skeletons inside SMPL mesh
 es in diverse poses. We build such a dataset by optimizing biomechanically
  accurate skeletons inside SMPL meshes from AMASS sequences. We then learn
  a regressor from SMPL mesh vertices to the true joint locations and bone 
 rotations. Finally, we re-parametrize the SMPL mesh with the new kinematic
  parameters.\nThe resulting SKEL model is animatable like SMPL but with fe
 wer, and biomechanically-realistic, degrees of freedom. We also train a re
 gressor from SMPL meshes to the skeleton enabling us to ``upgrade" existin
 g datasets that are in SMPL format. We show that SKEL has more biomechanic
 ally accurate joint locations than SMPL, and the bones fit inside the body
  surface better than previous methods.\nSKEL provides a new tool to enable
  biomechanics in the wild, while also providing vision and graphics resear
 chers with a better constrained and more realistic model of human articula
 tion.\n\nRegistration Category: Full Access\n\nSession Chair: Seungbae Ban
 g (Amazon)\n\n
URL:https://asia.siggraph.org/2023/full-program?id=papers_629&sess=sess166
END:VEVENT
END:VCALENDAR