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:20260817T171530Z
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:Simon Giebenhain and Tobias Kirschstein (Technical University 
 of Munich); Martin Rünz (Synthesia); Lourdes Agapito (University College L
 ondon (UCL), Synthesia); and Matthias Nießner (Technical University of Mun
 ich, Synthesia)\n\nThe creation of high-fidelity, digital versions of huma
 n heads is an important stepping stone in the process of further integrati
 ng virtual components into our everyday lives. Constructing such avatars i
 s a challenging research problem, due to a high demand for photo-realism a
 nd real-time rendering performance. In this work, we propose Neural Parame
 tric Gaussian Avatars (NPGA), a data-driven approach to create high-fideli
 ty, controllable avatars from multi-view video recordings. We build our me
 thod around 3D Gaussian splatting for its highly efficient rendering and t
 o inherit the topological flexibility of point clouds. In contrast to prev
 ious work, we condition our avatars’ dynamics on the rich expression space
  of neural parametric head models (NPHM), instead of mesh-based 3DMMs. To 
 this end, we distill the backward deformation field of our underlying NPHM
  into forward deformations which are compatible with rasterization-based r
 endering. All remaining fine-scale, expression-dependent details are learn
 ed from the multi-view videos. For increased representational capacity of 
 our avatars, we propose per-Gaussian latent features that condition each p
 rimitives dynamic behavior. To regularize this increased dynamic expressiv
 ity, we propose Laplacian terms on the latent features and predicted dynam
 ics. We evaluate our method on the public NeRSemble dataset, demonstrating
  that NPGA significantly outperforms the previous state-of-the-art avatars
  on the self-reenactment task by 2.6 PSNR. Furthermore, we demonstrate acc
 urate animation capabilities from real-world monocular videos.\n\nRegistra
 tion Category: Full Access, Full Access Supporter\n\nLanguage Format: Engl
 ish Language\n\nSession Chair: Iain Matthews (Epic Games, Carnegie Mellon 
 University)\n\n
URL:https://asia.siggraph.org/2024/program/?id=papers_1101&sess=sess143
END:VEVENT
END:VCALENDAR