BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Asia/Seoul
X-LIC-LOCATION:Asia/Seoul
BEGIN:STANDARD
TZOFFSETFROM:+0900
TZOFFSETTO:+0900
TZNAME:KST
DTSTART:18871231T000000
DTSTART:19881009T020000
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20230103T035307Z
LOCATION:Auditorium\, Level 5\, West Wing
DTSTART;TZID=Asia/Seoul:20221206T100000
DTEND;TZID=Asia/Seoul:20221206T120000
UID:siggraphasia_SIGGRAPH Asia 2022_sess153_papers_513@linklings.com
SUMMARY:NeuralRoom: Geometry-Constrained Neural Implicit Surfaces for Indo
 or Scene Reconstruction
DESCRIPTION:Technical Papers\n\nNeuralRoom: Geometry-Constrained Neural Im
 plicit Surfaces for Indoor Scene Reconstruction\n\nWang, Li, Jiang, Zhou, 
 Cao...\n\nWe present a novel neural surface reconstruction method called N
 euralRoom for reconstructing room-sized indoor scenes directly from a set 
 of 2D images. Recently, implicit neural representations have become a prom
 ising way to reconstruct surfaces from multiview images due to their high-
 quality results and simplicity. However, implicit neural representations u
 sually cannot reconstruct indoor scenes well because they suffer severe sh
 ape-radiance ambiguity. We assume that the indoor scene consists of textur
 e-rich and flat texture-less regions. In texture-rich regions, the multivi
 ew stereo can obtain accurate results. In the flat area, normal estimation
  networks usually obtain a good normal estimation. Based on the above obse
 rvations, we reduce the possible spatial variation range of implicit neura
 l surfaces by reliable geometric priors to alleviate shape-radiance ambigu
 ity. Specifically, we use multiview stereo results to limit the NeuralRoom
  optimization space and then use reliable geometric priors to guide Neural
 Room training. Then the NeuralRoom would produce a neural scene representa
 tion that can render an image consistent with the input training images. I
 n addition, we propose a smooth method called perturbation-residual restri
 ctions to improve the accuracy and completeness of the flat region, which 
 assumes that the sampling points in a local surface should have the same n
 ormal and similar distance to the observation center. Experiments on the S
 canNet dataset show that our method can reconstruct the texture-less area 
 of indoor scenes while maintaining the accuracy of detail. We also apply N
 euralRoom to more advanced multiview reconstruction algorithms and signifi
 cantly improve their reconstruction quality.\n\nRegistration Category: FUL
 L ACCESS, EXPERIENCE PLUS ACCESS, EXPERIENCE ACCESS, TRADE EXHIBITOR\n\nLa
 nguage: ENGLISH\n\nFormat: IN-PERSON
URL:https://sa2022.siggraph.org/en/full-program/?id=papers_513&sess=sess15
 3
END:VEVENT
END:VCALENDAR