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 (1)\, B Block\, Level 5
DTSTART;TZID=Asia/Tokyo:20241204T133400
DTEND;TZID=Asia/Tokyo:20241204T134600
UID:siggraphasia_SIGGRAPH Asia 2024_sess115_papers_126@linklings.com
SUMMARY:Real-time Large-scale Deformation of Gaussian Splatting
DESCRIPTION:Technical Papers\n\nLin Gao (Institute of Computing Technology
 , Chinese Academy of Sciences; University of Chinese Academy of Sciences);
  Jie Yang (Institute of Computing Technology, Chinese Academy of Sciences)
 ; Bo-Tao Zhang, Jia-Mu Sun, and Yu-Jie Yuan (Institute of Computing Techno
 logy, Chinese Academy of Sciences; University of Chinese Academy of Scienc
 es); Hongbo Fu (Hong Kong University of Science and Technology); and Yu-Ku
 n Lai (Cardiff University)\n\nNeural implicit representations, including N
 eural Distance Fields and Neural Radiance Fields, have demonstrated signif
 icant capabilities for reconstructing surfaces with complicated geometry a
 nd topology, and generating novel views of a scene. Nevertheless, it is ch
 allenging for users to directly deform or manipulate these implicit repres
 entations with large deformations in the real-time fashion.Gaussian Splatt
 ing (GS) has recently become a promising method with explicit geometry for
  representing static scenes and facilitating high-quality and real time sy
 nthesis of novel views. However, it cannot be easily deformed due to the u
 se of discrete Gaussians and lack of explicit topology. To address this, w
 e develop a novel GS-based method that enables interactive deformation. Ou
 r key idea is to design an innovative mesh-based GS representation, which 
 is integrated into Gaussian learning and manipulation. 3D Gaussians are de
 fined over an explicit mesh, and they are bound with each other: the rende
 ring of 3D Gaussians guides the mesh face split for adaptive refinement, a
 nd the mesh face split directs the splitting of 3D Gaussians. Moreover, th
 e explicit mesh constraints help regularize the Gaussian distribution, sup
 pressing poor-quality Gaussians (e.g. , misaligned Gaussians, long-narrow 
 shaped Gaussians), thus enhancing visual quality and avoiding artifacts du
 ring deformation. Based on this representation, we further introduce a lar
 ge-scale Gaussian deformation technique to enable deformable GS, which alt
 ers the parameters of 3D Gaussians according to the manipulation of the as
 sociated mesh. Our method benefits from existing mesh deformation datasets
  for more realistic data-driven Gaussian deformation. Extensive experiment
 s show that our approach achieves high-quality reconstruction and effectiv
 e deformation, while maintaining promising rendering results at a high fra
 me rate (65 FPS on average on a single commodity GPU).\n\nRegistration Cat
 egory: Full Access, Full Access Supporter\n\nLanguage Format: English Lang
 uage\n\nSession Chair: Peng-Shuai Wang (Peking University)
URL:https://asia.siggraph.org/2024/program/?id=papers_126&sess=sess115
END:VEVENT
END:VCALENDAR