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:20250110T023309Z
LOCATION:Hall B5 (2)\, B Block\, Level 5
DTSTART;TZID=Asia/Tokyo:20241203T132800
DTEND;TZID=Asia/Tokyo:20241203T134200
UID:siggraphasia_SIGGRAPH Asia 2024_sess104_papers_415@linklings.com
SUMMARY:Representing Long Volumetric Video with Temporal Gaussian Hierarch
 y
DESCRIPTION:Technical Papers\n\nZhen Xu (State Key Laboratory of CAD&CG, Z
 hejiang University; Zhejiang University); Yinghao Xu (Stanford University)
 ; Zhiyuan Yu (Department of Mathematics, Hong Kong University of Science a
 nd Technology); Sida Peng and Jiaming Sun (Zhejiang University); and Hujun
  Bao and Xiaowei Zhou (State Key Laboratory of CAD&CG, Zhejiang University
 )\n\nThis paper aims to address the challenge of reconstructing long volum
 etric videos from multi-view RGB videos.\nRecent dynamic view synthesis me
 thods leverage powerful 4D representations, like feature grids or point cl
 oud sequences, to achieve high-quality rendering results. However, they ar
 e typically limited to short (1$\sim$2s) video clips and often suffer from
  large memory footprints when dealing with longer videos.\nTo solve this i
 ssue, we propose a novel 4D representation, named temporal Gaussian hierar
 chy, to compactly model long volumetric videos.\nOur key observation is th
 at there are generally various degrees of temporal redundancy in dynamic s
 cenes, which consist of areas changing at different speeds.\nMotivated by 
 this, our approach builds a multi-level hierarchy of Gaussian primitives, 
 where each level separately describes scene regions with different degrees
  of content change, and adaptively shares Gaussian primitives to represent
  unchanged scene content over different temporal segments, thus effectivel
 y reducing the number of Gaussian primitives.\nIn addition, the tree-like 
 structure of the Gaussian hierarchy allows us to efficiently represent the
  scene at a particular moment with a subset of Gaussian primitives, leadin
 g to nearly constant GPU memory usage during the training or rendering reg
 ardless of the video length.\nMoreover, we design a compact appearance mod
 el that mixes diffuse and view-dependent Gaussians to further minimize the
  model size while maintaining the rendering quality.\nWe also develop a ra
 sterization pipeline of Gaussian primitives based on the hardware-accelera
 ted technique to improve rendering speed.\nExtensive experimental results 
 demonstrate the superiority of our method over alternative methods in term
 s of training cost, rendering speed, and storage usage.\nTo our knowledge,
  this work is the first approach capable of efficiently handling hours of 
 volumetric video data while maintaining state-of-the-art rendering quality
 .\n\nRegistration Category: Full Access, Full Access Supporter\n\nLanguage
  Format: English Language\n\nSession Chair: Bernhard Kerbl (Technical Univ
 ersity of Vienna)
URL:https://asia.siggraph.org/2024/program/?id=papers_415&sess=sess104
END:VEVENT
END:VCALENDAR