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:20260817T171530Z
LOCATION:Hall B5 (1)\, B Block\, Level 5
DTSTART;TZID=Asia/Tokyo:20241204T131100
DTEND;TZID=Asia/Tokyo:20241204T132300
UID:siggraphasia_SIGGRAPH Asia 2024_sess115_papers_678@linklings.com
SUMMARY:3DGSR: Implicit Surface Reconstruction with 3D Gaussian Splatting
DESCRIPTION:Xiaoyang Lyu, Yang-Tian Sun, Yi-Hua Huang, Xiuzhe Wu, and Ziyi
  Yang (University of Hong Kong); Yilun Chen and Jiangmiao Pang (Shanghai A
 rtificial Intelligence Laboratory); and Xiaojuan Qi (University of Hong Ko
 ng)\n\nIn this paper, we present an implicit surface reconstruction method
  with\n3D Gaussian Splatting (3DGS), namely 3DGSR, that allows for accurat
 e 3D\nreconstruction with intricate details while inheriting the high effi
 ciency and\nrendering quality of 3DGS. The key insight is to incorporate a
 n implicit\nsigned distance field (SDF) within 3D Gaussians for surface mo
 deling and to\nenable the alignment and joint optimization of both SDF and
  3D Gaussians.\nTo achieve this, we design coupling strategies that align 
 and associate the\nSDF with 3D Gaussians, allowing for unified optimizatio
 n and enforcing\nsurface constraints on the 3D Gaussians. With alignment, 
 optimizing the 3D\nGaussians provides supervisory signals for SDF learning
 , enabling the recon-\nstruction of intricate details. However, this only 
 offers sparse supervisory\nsignals to the SDF at locations occupied by Gau
 ssians, which is insufficient\nfor learning a continuous SDF. Then, to add
 ress this limitation, we incor-\nporate volumetric rendering and align the
  rendered geometric attributes\n(depth, normal) with that derived from 3DG
 S. In sum, these two designs\nallow SDF and 3DGS to be aligned, jointly op
 timized, and mutually boosted.\nOur extensive experimental results demonst
 rate that our 3DGSR enables\nhigh-quality 3D surface reconstruction while 
 preserving the efficiency and\nrendering quality of 3DGS. Besides, our met
 hod competes favorably with\nleading surface reconstruction techniques whi
 le offering a more efficient\nlearning process and much better rendering q
 ualities.\n\nRegistration Category: Full Access, Full Access Supporter\n\n
 Language Format: English Language\n\nSession Chair: Peng-Shuai Wang (Pekin
 g University)\n\n
URL:https://asia.siggraph.org/2024/program/?id=papers_678&sess=sess115
END:VEVENT
END:VCALENDAR