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:20260817T171532Z
LOCATION:Hall B7 (1)\, B Block\, Level 7
DTSTART;TZID=Asia/Tokyo:20241205T144500
DTEND;TZID=Asia/Tokyo:20241205T145600
UID:siggraphasia_SIGGRAPH Asia 2024_sess135_papers_156@linklings.com
SUMMARY:ToonCrafter: Generative Cartoon Interpolation
DESCRIPTION:Jinbo Xing (Chinese University of Hong Kong); Hanyuan Liu (Cit
 y University of Hong Kong); Menghan Xia, Yong Zhang, Xintao Wang, and Ying
  Shan (Tencent); and Tien-Tsin Wong (Monash University, Chinese University
  of Hong Kong)\n\nWe introduce ToonCrafter, a novel approach that transcen
 ds traditional correspondence-based cartoon video interpolation, paving th
 e way for generative interpolation. Traditional methods, that implicitly a
 ssume linear motion and the absence of complicated phenomena like dis-occl
 usion, often struggle with the exaggerated non-linear and large motions wi
 th occlusion commonly found in cartoons, resulting in implausible or even 
 failed interpolation results. To overcome these limitations, we explore th
 e potential of adapting live-action video priors to better suit cartoon in
 terpolation within a generative framework. ToonCrafter effectively address
 es the challenges faced when applying live-action video motion priors to g
 enerative cartoon interpolation. First, we design a toon rectification lea
 rning strategy that seamlessly adapts live-action video priors to the cart
 oon domain, resolving the domain gap  and content leakage issues. Next, we
  introduce a dual-reference-based 3D decoder to compensate for lost detail
 s due to the highly compressed latent prior spaces, ensuring the preservat
 ion of fine details in interpolation results. Finally, we design a flexibl
 e sketch encoder that empowers users with interactive control over the int
 erpolation results. Experimental results demonstrate that our proposed met
 hod not only produces visually convincing and more natural dynamics, but a
 lso effectively handles dis-occlusion. The comparative evaluation demonstr
 ates the notable superiority of our approach over existing competitors.\n\
 nRegistration Category: Full Access, Full Access Supporter\n\nLanguage For
 mat: English Language\n\nSession Chair: Changjian Li (University of Edinbu
 rgh)\n\n
URL:https://asia.siggraph.org/2024/program/?id=papers_156&sess=sess135
END:VEVENT
END:VCALENDAR