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:20250110T023313Z
LOCATION:Hall B5 (1)\, B Block\, Level 5
DTSTART;TZID=Asia/Tokyo:20241206T113100
DTEND;TZID=Asia/Tokyo:20241206T114300
UID:siggraphasia_SIGGRAPH Asia 2024_sess142_papers_700@linklings.com
SUMMARY:Style-NeRF2NeRF: 3D Style Transfer from Style-Aligned Multi-View I
 mages
DESCRIPTION:Technical Papers\n\nHaruo Fujiwara (University of Tokyo) and Y
 usuke Mukuta and Tatsuya Harada (University of Tokyo, RIKEN AIP)\n\nWe pro
 pose a simple yet effective pipeline for stylizing a 3D scene, harnessing 
 the power of 2D image diffusion models. Given a NeRF model reconstructed f
 rom a set of multi-view images, we perform 3D style transfer by refining t
 he source NeRF model using stylized images generated by a style-aligned im
 age-to-image diffusion model.\nGiven a target style prompt, we first gener
 ate perceptually similar multi-view images by leveraging a depth-condition
 ed diffusion model with an attention-sharing mechanism. Next, based on the
  stylized multi-view images, we propose to guide the style transfer proces
 s with the sliced Wasserstein loss based on the feature maps extracted fro
 m a pre-trained CNN model.\nOur pipeline consists of decoupled steps, allo
 wing users to test various prompt ideas and preview the stylized 3D result
  before proceeding to the NeRF fine-tuning stage.\nWe demonstrate that our
  method can transfer diverse artistic styles to real-world 3D scenes with 
 competitive quality.\n\nRegistration Category: Full Access, Full Access Su
 pporter\n\nLanguage Format: English Language\n\nSession Chair: Maria Larss
 on (University of Tokyo)
URL:https://asia.siggraph.org/2024/program/?id=papers_700&sess=sess142
END:VEVENT
END:VCALENDAR