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:20241204T135800
DTEND;TZID=Asia/Tokyo:20241204T140900
UID:siggraphasia_SIGGRAPH Asia 2024_sess115_papers_1225@linklings.com
SUMMARY:L3DG: Latent 3D Gaussian Diffusion
DESCRIPTION:Barbara Roessle (Technical University of Munich); Norman Mülle
 r, Lorenzo Porzi, Samuel Rota Bulò, and Peter Kontschieder (Meta Reality L
 abs); and Angela Dai and Matthias Nießner (Technical University of Munich)
 \n\nWe propose L3DG, the first approach for generative 3D modeling of 3D G
 aussians through a latent 3D Gaussian diffusion formulation.\nThis enables
  effective generative 3D modeling, scaling to generation of entire room-sc
 ale scenes which can be very efficiently rendered.\nTo enable effective sy
 nthesis of 3D Gaussians, we propose a latent diffusion formulation, operat
 ing in a compressed latent space of 3D Gaussians.\nThis compressed latent 
 space is learned by a vector-quantized variational autoencoder (VQ-VAE), f
 or which we employ a sparse convolutional architecture to efficiently oper
 ate on room-scale scenes. \nThis way, the complexity of the costly generat
 ion process via diffusion is substantially reduced, allowing higher detail
  on object-level generation, as well as scalability to large scenes. \nBy 
 leveraging the 3D Gaussian representation, the generated scenes can be ren
 dered from arbitrary viewpoints in real-time. \nWe demonstrate that our ap
 proach significantly improves visual quality over prior work on unconditio
 nal object-level radiance field synthesis and showcase its applicability t
 o room-scale scene generation.\n\nRegistration Category: Full Access, Full
  Access Supporter\n\nLanguage Format: English Language\n\nSession Chair: P
 eng-Shuai Wang (Peking University)\n\n
URL:https://asia.siggraph.org/2024/program/?id=papers_1225&sess=sess115
END:VEVENT
END:VCALENDAR