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:20250110T023313Z
LOCATION:Hall B5 (1)\, B Block\, Level 5
DTSTART;TZID=Asia/Tokyo:20241206T111900
DTEND;TZID=Asia/Tokyo:20241206T113100
UID:siggraphasia_SIGGRAPH Asia 2024_sess142_papers_205@linklings.com
SUMMARY:DreamUDF: Generating Unsigned Distance Fields from A Single Image
DESCRIPTION:Technical Papers\n\nYu-Tao Liu and Xuan Gao (Institute of Comp
 uting Technology, Chinese Academy of Sciences; University of Chinese Acade
 my of Sciences); Weikai Chen (Tencent Games); Jie Yang (Institute of Compu
 ting Technology, Chinese Academy of Sciences; University of Chinese Academ
 y of Sciences); Xiaoxu Meng and Bo Yang (Tencent Games); and Lin Gao (Inst
 itute of Computing Technology, Chinese Academy of Sciences; University of 
 Chinese Academy of Sciences)\n\nRecent advances in diffusion models and ne
 ural implicit surfaces have shown promising progress in generating 3D mode
 ls. However, existing generative frameworks are limited to closed surfaces
 , failing to cope with a wide range of commonly seen shapes that have open
  boundaries. In this work, we present DreamUDF, a novel framework for gene
 rating high-quality 3D objects with arbitrary topologies from a single ima
 ge. To address the challenge of generating proper topology given sparse an
 d ambiguous observations, we propose to incorporate both the data priors f
 rom a multi-view diffusion model and the geometry priors brought by an uns
 iged distance field (UDF) reconstructor. In particular, we leverage a join
 t framework that consists of 1) a generative module that produces a neural
  radiance field that provides photo-realistic renderings from the arbitrar
 y view; and 2) a reconstructive module that distills the learnable radianc
 e field into surfaces with arbitrary topologies. We further introduce a fi
 eld coupler that bridges the radiance field and UDF under an novel optimiz
 ation scheme. This allows the two modules to mutually boost each other dur
 ing training. Extensive experiments and evaluations demonstrate that Dream
 UDF achieves high-quality reconstruction and robust 3D generation on both 
 closed and open surfaces with arbitrary topologies, compared to the previo
 us works.\n\nRegistration Category: Full Access, Full Access Supporter\n\n
 Language Format: English Language\n\nSession Chair: Maria Larsson (Univers
 ity of Tokyo)
URL:https://asia.siggraph.org/2024/program/?id=papers_205&sess=sess142
END:VEVENT
END:VCALENDAR