BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Asia/Hong_Kong
X-LIC-LOCATION:Asia/Hong_Kong
BEGIN:STANDARD
TZOFFSETFROM:+0800
TZOFFSETTO:+0800
TZNAME:HKT
DTSTART:19911015T033000
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20251218T030656Z
LOCATION:Meeting Room S426+S427\, Level 4
DTSTART;TZID=Asia/Hong_Kong:20251215T131000
DTEND;TZID=Asia/Hong_Kong:20251215T132000
UID:siggraphasia_SIGGRAPH Asia 2025_sess106_papers_1079@linklings.com
SUMMARY:Vertex Features for Neural Global Illumination
DESCRIPTION:Rui Su, Honghao Dong, Haojie Jin, Yisong Chen, Guoping Wang, a
 nd Sheng Li (Peking University)\n\nRecent research on learnable neural rep
 resentations has been widely adopted in the field of 3D scene reconstructi
 on and neural rendering applications. However, traditional feature grid re
 presentations often suffer from substantial memory footprint, posing a sig
 nificant bottleneck for modern parallel computing hardware. In this paper,
  we present neural vertex features, a generalized formulation of learnable
  representation for neural rendering tasks involving explicit mesh surface
 s. Instead of uniformly distributing neural features throughout 3D space, 
 our method stores learnable features directly at mesh vertices, leveraging
  the underlying geometry as a compact and structured representation for ne
 ural processing. This not only optimizes memory efficiency, but also impro
 ves feature representation by aligning compactly with the surface using ta
 sk-specific geometric priors. Additionally, neural vertex features offer i
 mproved feature representation by compactly aligning with the surface usin
 g task-specific geometric priors. We validate our neural representation ac
 ross diverse neural rendering tasks, with a specific emphasis on neural ra
 diosity. Experimental results demonstrate that our method reduces memory c
 onsumption to only one-fifth (or even less) of grid-based representations,
  while maintaining comparable rendering quality and lowering inference ove
 rhead.\n\nRegistration Category: Full Access, Full Access Supporter\n\nSes
 sion Chair: Yuchi Huo (Zhejiang University, Korea Advanced Institute of Sc
 ience and Technology)\n\n
END:VEVENT
END:VCALENDAR