BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Australia/Melbourne
X-LIC-LOCATION:Australia/Melbourne
BEGIN:DAYLIGHT
TZOFFSETFROM:+1000
TZOFFSETTO:+1100
TZNAME:AEDT
DTSTART:19721003T020000
RRULE:FREQ=YEARLY;BYMONTH=4;BYDAY=1SU
END:DAYLIGHT
BEGIN:STANDARD
DTSTART:19721003T020000
TZOFFSETFROM:+1100
TZOFFSETTO:+1000
TZNAME:AEST
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20240214T070242Z
LOCATION:Meeting Room C4.8\, Level 4 (Convention Centre)
DTSTART;TZID=Australia/Melbourne:20231212T160000
DTEND;TZID=Australia/Melbourne:20231212T161000
UID:siggraphasia_SIGGRAPH Asia 2023_sess160_papers_998@linklings.com
SUMMARY:MOCHA: Real-Time Motion Characterization via Context Matching
DESCRIPTION:Technical Papers, TOG\n\nDeok-Kyeong Jang (KAIST, MOVIN Inc.);
  Yuting Ye (Meta); Jungdam Won (Seoul National University); and Sung-Hee L
 ee (KAIST)\n\nTransforming neutral, characterless input motions to embody 
 the distinct style of a notable character in real time is highly compellin
 g for character animation. This paper introduces MOCHA, a novel online mot
 ion characterization framework that transfers both motion styles and body 
 proportions from a target character to an input source motion. MOCHA begin
 s by encoding the input motion into a motion feature that structures the b
 ody part topology and captures motion dependencies for effective character
 ization. Central to our framework is the Neural Context Matcher, which gen
 erates a motion feature for the target character with the most similar con
 text to the input motion feature. The conditioned autoregressive model of 
 the Neural Context Matcher can produce temporally coherent character featu
 res in each time frame. To generate the final characterized pose, our Char
 acterizer network incorporates the characteristic aspects of the target mo
 tion feature into the input motion feature while preserving its context. T
 his is achieved through a transformer model that introduces the adaptive i
 nstance normalization and context mapping-based cross-attention, effective
 ly injecting the character feature into the source feature. We validate th
 e performance of our framework through comparisons with prior work and an 
 ablation study. Our framework can easily accommodate various applications,
  including characterization with only sparse input and real-time character
 ization. Additionally, we contribute a high-quality motion dataset compris
 ing six different characters performing a range of motions, which can serv
 e as a valuable resource for future research.\n\nRegistration Category: Fu
 ll Access\n\nSession Chair: Ioannis Karamouzas (Clemson University, Univer
 sity of California Riverside)
URL:https://asia.siggraph.org/2023/full-program?id=papers_998&sess=sess160
END:VEVENT
END:VCALENDAR