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:20240214T070249Z
LOCATION:Meeting Room C4.8\, Level 4 (Convention Centre)
DTSTART;TZID=Australia/Melbourne:20231215T103000
DTEND;TZID=Australia/Melbourne:20231215T104500
UID:siggraphasia_SIGGRAPH Asia 2023_sess154_papers_485@linklings.com
SUMMARY:ScaNeRF: Scalable Bundle-Adjusting Neural Radiance Fields for Larg
 e-Scale Scene Rendering
DESCRIPTION:Technical Papers\n\nXiuchao Wu (State Key Laboratory of CAD & 
 CG, Zhejiang University); Jiamin Xu (Hangzhou Dianzi Univeristy); Xin Zhan
 g (State Key Laboratory of CAD&CG, Zhejiang Univerisity); Hujun Bao (State
  Key Laboratory of CAD & CG, Zhejiang University); Qixing Huang (Universit
 y of Texas at Austin); Yujun Shen (Ant Group); James Tompkin (Brown Univer
 sity); and Weiwei Xu (State Key Laboratory of CAD&CG, Zhejiang Univerisity
 )\n\nHigh-quality large-scale scene rendering requires a scalable represen
 tation and accurate camera poses. This research combines tile-based hybrid
  neural fields with parallel distributive optimization to improve bundle-a
 djusting neural radiance fields. The proposed method scales with a divide-
 and-conquer strategy. We partition scenes into tiles, each with a multi-re
 solution hash feature grid and shallow chained diffuse and specular multi-
 layer perceptrons (MLPs). Tiles unify foreground and background via a spat
 ial contraction function that allows both distant objects in outdoor scene
 s and planar reflections as virtual images outside the tile. Decomposing a
 ppearance with the specular MLP allows a specular-aware warping loss to pr
 ovide a second optimization path for camera poses. We apply the alternatin
 g direction method of multipliers (ADMM) to achieve consensus among camera
  poses while maintaining parallel tile optimization. Experimental results 
 show that our method outperforms state-of-the-art neural scene rendering m
 ethod quality by 5%--10% in PSNR, maintaining sharp distant objects and vi
 ew-dependent reflections across six indoor and outdoor scenes.\n\nRegistra
 tion Category: Full Access\n\nSession Chair: Yuchi Huo (Zhejiang Universit
 y, Korea Advanced Institute of Science and Technology)
URL:https://asia.siggraph.org/2023/full-program?id=papers_485&sess=sess154
END:VEVENT
END:VCALENDAR