BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Asia/Tokyo
X-LIC-LOCATION:Asia/Tokyo
BEGIN:STANDARD
TZOFFSETFROM:+0900
TZOFFSETTO:+0900
TZNAME:JST
DTSTART:18871231T000000
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20250110T023313Z
LOCATION:Hall B5 (2)\, B Block\, Level 5
DTSTART;TZID=Asia/Tokyo:20241206T113100
DTEND;TZID=Asia/Tokyo:20241206T114300
UID:siggraphasia_SIGGRAPH Asia 2024_sess143_papers_1101@linklings.com
SUMMARY:NPGA: Neural Parametric Gaussian Avatars
DESCRIPTION:Technical Papers\n\nSimon Giebenhain and Tobias Kirschstein (T
 echnical University of Munich); Martin Rünz (Synthesia); Lourdes Agapito (
 University College London (UCL), Synthesia); and Matthias Nießner (Technic
 al University of Munich, Synthesia)\n\nThe creation of high-fidelity, digi
 tal versions of human heads is an important stepping stone in the process 
 of further integrating virtual components into our everyday lives. Constru
 cting such avatars is a challenging research problem, due to a high demand
  for photo-realism and real-time rendering performance. In this work, we p
 ropose Neural Parametric Gaussian Avatars (NPGA), a data-driven approach t
 o create high-fidelity, controllable avatars from multi-view video recordi
 ngs. We build our method around 3D Gaussian splatting for its highly effic
 ient rendering and to inherit the topological flexibility of point clouds.
  In contrast to previous work, we condition our avatars’ dynamics on the r
 ich expression space of neural parametric head models (NPHM), instead of m
 esh-based 3DMMs. To this end, we distill the backward deformation field of
  our underlying NPHM into forward deformations which are compatible with r
 asterization-based rendering. All remaining fine-scale, expression-depende
 nt details are learned from the multi-view videos. For increased represent
 ational capacity of our avatars, we propose per-Gaussian latent features t
 hat condition each primitives dynamic behavior. To regularize this increas
 ed dynamic expressivity, we propose Laplacian terms on the latent features
  and predicted dynamics. We evaluate our method on the public NeRSemble da
 taset, demonstrating that NPGA significantly outperforms the previous stat
 e-of-the-art avatars on the self-reenactment task by 2.6 PSNR. Furthermore
 , we demonstrate accurate animation capabilities from real-world monocular
  videos.\n\nRegistration Category: Full Access, Full Access Supporter\n\nL
 anguage Format: English Language\n\nSession Chair: Iain Matthews (Epic Gam
 es, Carnegie Mellon University)
URL:https://asia.siggraph.org/2024/program/?id=papers_1101&sess=sess143
END:VEVENT
END:VCALENDAR