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:20260817T171531Z
LOCATION:Hall B5 (2)\, B Block\, Level 5
DTSTART;TZID=Asia/Tokyo:20241203T171600
DTEND;TZID=Asia/Tokyo:20241203T172800
UID:siggraphasia_SIGGRAPH Asia 2024_sess110_papers_1049@linklings.com
SUMMARY:Dynamic Neural Radiosity with Multi-grid Decomposition
DESCRIPTION:Rui Su, Honghao Dong, Jierui Ren, Haojie Jin, Yisong Chen, Guo
 ping Wang, and Sheng Li (Peking University)\n\nPrior approaches to the neu
 ral rendering of global illumination typically rely on complex network arc
 hitectures and training strategies to model the global effects. This often
  leads to impractically high overheads for both training and inference. Th
 e neural radiosity technique marks a significant advancement by injecting 
 the radiometric prior into the training process, allowing for efficient mo
 deling of the global radiance fields using a lightweight network and grid-
 based representations. However, this method encounters difficulties in mod
 eling dynamic scenes, as the high-dimensional feature space quickly become
 s unmanageable as the number of varying scene parameters grows. In this wo
 rk, we extend neural radiosity for variable scenes through a novel neural 
 decomposition method. To achieve this, we first parameterize the animated 
 scene with an explicit vector $\mathbf{v}$, which conditions a high-dimens
 ional radiance field $L_{\theta}$. We then develop a practical representat
 ion for $L_{\theta}$ by decomposing the high-dimensional feature grid into
  3D grids, 2D feature planes, and lightweight MLPs. This strategy effectiv
 ely models the correlation between 3D spatial features and dynamic scene v
 ariables, while maintaining a practical memory and computational cost. Exp
 erimental results show that our method facilitates efficient dynamic globa
 l illumination rendering with practical runtime performance, outperforming
  previous state-of-the-art techniques with both reduced training and infer
 ence costs.\n\nRegistration Category: Full Access, Full Access Supporter\n
 \nLanguage Format: English Language\n\nSession Chair: Michael Wimmer (TU W
 ien, Technische Universität Wien (TU Wien))\n\n
URL:https://asia.siggraph.org/2024/program/?id=papers_1049&sess=sess110
END:VEVENT
END:VCALENDAR