BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Asia/Seoul
X-LIC-LOCATION:Asia/Seoul
BEGIN:STANDARD
TZOFFSETFROM:+0900
TZOFFSETTO:+0900
TZNAME:KST
DTSTART:18871231T000000
DTSTART:19881009T020000
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20230103T035309Z
LOCATION:Room 324\, Level 3\, West Wing
DTSTART;TZID=Asia/Seoul:20221207T140000
DTEND;TZID=Asia/Seoul:20221207T153000
UID:siggraphasia_SIGGRAPH Asia 2022_sess159_papers_436@linklings.com
SUMMARY:Reference Based Sketch Extraction via Attention Mechanism
DESCRIPTION:Technical Communications, Technical Papers\n\nReference Based 
 Sketch Extraction via Attention Mechanism\n\nAshtari, Seo, Kang, Cha, Noh\
 n\nWe propose a model that extracts a sketch from a colorized image in suc
 h a way that the extracted sketch has a line style similar to a given refe
 rence sketch while preserving the visual content identically to the colori
 zed image. Authentic sketches drawn by artists have various sketch styles 
 to add visual interest and contribute feeling to the sketch. However, exis
 ting sketch-extraction methods generate sketches with only one style. More
 over, existing style transfer models fail to transfer sketch styles becaus
 e they are mostly designed to transfer textures of a source style image in
 stead of transferring the sparse line styles from a reference sketch. Lack
 ing the necessary volumes of data for standard training of translation sys
 tems, at the core of our GAN-based solution is a self-reference sketch sty
 le generator that produces various reference sketches with a similar style
  but different spatial layouts. We use independent attention modules to de
 tect the edges of a colorized image and reference sketch as well as the vi
 sual correspondences between them. We apply several loss terms to imitate 
 the style and enforce sparsity in the extracted sketches. Our sketch-extra
 ction method results in a close imitation of a reference sketch style draw
 n by an artist and outperforms all baseline methods. Using our method, we 
 produce a synthetic dataset representing various sketch styles and improve
  the performance of auto-colorization models, in high demand in comics. Th
 e validity of our approach is confirmed via qualitative and quantitative e
 valuations.\n\nRegistration Category: FULL ACCESS, ON-DEMAND ACCESS\n\nLan
 guage: ENGLISH\n\nFormat: IN-PERSON, ON-DEMAND
URL:https://sa2022.siggraph.org/en/full-program/?id=papers_436&sess=sess15
 9
END:VEVENT
END:VCALENDAR