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:20260114T163642Z
LOCATION:Meeting Room C4.8\, Level 4 (Convention Centre)
DTSTART;TZID=Australia/Melbourne:20231212T172000
DTEND;TZID=Australia/Melbourne:20231212T173000
UID:siggraphasia_SIGGRAPH Asia 2023_sess142_papers_652@linklings.com
SUMMARY:Adaptive Recurrent Frame Prediction with Learnable Motion Vectors
DESCRIPTION:Zhizhen Wu (State Key Lab of CAD&CG, Zhejiang University); Che
 nyu Zuo (State Key Lab of CAD&CG, State Key Laboratory of CAD & CG, Zhejia
 ng University); Yuchi Huo (State Key Lab of CAD&CG, Zhejiang University; Z
 hejiang Lab); Yazhen Yuan (Tencent); Yifan Peng (The University of Hong Ko
 ng (HKU)); Guiyang Pu (China Mobile (Hangzhou) Information Technology Co.,
  Ltd); and Rui Wang and Hujun Bao (State Key Lab of CAD&CG, Zhejiang Unive
 rsity)\n\nThe utilization of dedicated ray tracing graphics cards has cont
 ributed to the production of stunning visual effects in real-time renderin
 g. However, the demand for high frame rates and high resolutions remains a
  challenge to be addressed. A crucial technique for increasing frame rate 
 and resolution is the pixel warping approach, which exploits spatio-tempor
 al coherence. \nTo this end, existing super-resolution and frame predictio
 n methods rely heavily on motion vectors from rendering engine pipelines t
 o track object movements. \nThis work builds upon state-of-the-art heurist
 ic approaches by exploring a novel adaptive recurrent frame prediction fra
 mework that integrates learnable motion vectors. Our framework supports th
 e prediction of transparency, particles, and texture animations, with impr
 oved motion vectors that capture shading, reflections, and occlusions, in 
 addition to geometry movements. \nWe also introduce a feature streaming ne
 ural network, dubbed FSNet, that allows for the adaptive prediction of one
  or multiple sequential frames. Extensive experiments against state-of-the
 -art methods demonstrate that FSNet can operate at lower latency with sign
 ificant visual enhancements and can upscale frame rates by at least two ti
 mes. This approach offers a flexible pipeline to improve the rendering fra
 me rates of various graphics applications and devices.\n\nRegistration Cat
 egory: Full Access\n\nSession Chair: Michael Gharbi (Reve AI, Massachusett
 s Institute of Technology (MIT))\n\n
URL:https://asia.siggraph.org/2023/full-program?id=papers_652&sess=sess142
END:VEVENT
END:VCALENDAR