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 B7 (1)\, B Block\, Level 7
DTSTART;TZID=Asia/Tokyo:20241203T153100
DTEND;TZID=Asia/Tokyo:20241203T154300
UID:siggraphasia_SIGGRAPH Asia 2024_sess108_papers_373@linklings.com
SUMMARY:LVCD: Reference-based Lineart Video Colorization with Diffusion Mo
 dels
DESCRIPTION:Zhitong Huang (City University of Hong Kong); Mohan Zhang (Wec
 hat, Tencent Inc.); and Jing Liao (City University of Hong Kong)\n\nWe pro
 pose the first video diffusion framework for reference-based lineart video
  colorization. Unlike previous works that rely solely on image generative 
 models to colorize lineart frame by frame, our approach leverages a large-
 scale pretrained video diffusion model to generate colorized animation vid
 eos. This approach leads to more temporally consistent results and is bett
 er equipped to handle large motions. Firstly, we introduce Sketch-guided C
 ontrolNet which provides additional control to finetune an image-to-video 
 diffusion model for controllable video synthesis, enabling the generation 
 of animation videos conditioned on lineart. We then propose Reference Atte
 ntion to facilitate the transfer of colors from the reference frame to oth
 er frames containing fast and expansive motions. Finally, we present a nov
 el scheme for sequential sampling, incorporating the Overlapped Blending M
 odule and Prev-Reference Attention, to extend the video diffusion model be
 yond its original fixed-length limitation for long video colorization. Bot
 h qualitative and quantitative results demonstrate that our method signifi
 cantly outperforms state-of-the-art techniques in terms of frame and video
  quality, as well as temporal consistency. Moreover, our method is capable
  of generating high-quality, long temporal-consistent animation videos wit
 h large motions, which is not achievable in previous works. Our code and m
 odel are available at https://luckyhzt.github.io/lvcd.\n\nRegistration Cat
 egory: Full Access, Full Access Supporter\n\nLanguage Format: English Lang
 uage\n\nSession Chair: I-Chao Shen (The University of Tokyo)\n\n
URL:https://asia.siggraph.org/2024/program/?id=papers_373&sess=sess108
END:VEVENT
END:VCALENDAR