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:20250110T023312Z
LOCATION:Hall B5 (2)\, B Block\, Level 5
DTSTART;TZID=Asia/Tokyo:20241205T134600
DTEND;TZID=Asia/Tokyo:20241205T135800
UID:siggraphasia_SIGGRAPH Asia 2024_sess131_papers_805@linklings.com
SUMMARY:Manifold Sampling for Differentiable Uncertainty in Radiance Field
 s
DESCRIPTION:Technical Papers\n\nLinjie Lyu (Max Planck Institute for Infor
 matics), Ayush Tewari (MIT CSAIL), Marc Habermann (Max Planck Institute fo
 r Informatics), Shunsuke Saito and Michael Zollhöfer (Meta Codec Avatars L
 ab), and Thomas Leimkühler and Christian Theobalt (Max Planck Institute fo
 r Informatics)\n\nRadiance fields are powerful and, hence, popular models 
 for representing the appearance of complex scenes. Yet, constructing them 
 based on image observations gives rise to ambiguities and uncertainties. W
 e propose a versatile approach for learning Gaussian radiance fields with 
 explicit and fine-grained uncertainty estimates that impose only little ad
 ditional cost compared to uncertainty-agnostic training. Our key observati
 on is that uncertainties can be modeled as a low-dimensional manifold in t
 he space of radiance field parameters that is highly amenable to Monte Car
 lo sampling. Importantly, our uncertainties are differentiable and, thus, 
 allow for gradient-based optimization of subsequent captures that optimall
 y reduce ambiguities. We demonstrate state-of-the-art performance on next-
 best-view planning tasks, including high-dimensional illumination planning
  for optimal radiance field relighting quality.\n\nRegistration Category: 
 Full Access, Full Access Supporter\n\nLanguage Format: English Language\n\
 nSession Chair: Seungyong Lee (POSTECH)
URL:https://asia.siggraph.org/2024/program/?id=papers_805&sess=sess131
END:VEVENT
END:VCALENDAR