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:20260114T163631Z
LOCATION:Darling Harbour Theatre\, Level 2 (Convention Centre)
DTSTART;TZID=Australia/Melbourne:20231212T093000
DTEND;TZID=Australia/Melbourne:20231212T124500
UID:siggraphasia_SIGGRAPH Asia 2023_sess209_papers_672@linklings.com
SUMMARY:Inovis: Instant Novel-View Synthesis
DESCRIPTION:Mathias Harrer and Linus Franke (Friedrich-Alexander-Universit
 ät Erlangen-Nürnberg); Laura Fink (Friedrich-Alexander-Universität Erlange
 n-Nürnberg, Fraunhofer IIS); and Marc Stamminger and Tim Weyrich (Friedric
 h-Alexander-Universität Erlangen-Nürnberg)\n\nNovel-view synthesis is an i
 ll-posed problem in that it requires inference of previously unseen inform
 ation. Recently, reviving the traditional field of image-based rendering, 
 neural methods proved particularly suitable for this interpolation/extrapo
 lation task; however, they often require a-priori scene-completeness or co
 stly pre-processing steps and generally suffer from long (scene-specific) 
 training times. Our work draws from recent progress in neural spatio-tempo
 ral supersampling to enhance a state-of-the-art neural renderer’s ability 
 to infer novel-view information at inference time. We adapt a supersamplin
 g architecture [Xiao et al. 2020], which resamples previously rendered fra
 mes, to instead recombine nearby camera images in a multi-view dataset. Th
 ese input frames are warped into a joint target frame, guided by the most 
 recent (point-based) scene representation, followed by neural interpolatio
 n. The resulting architecture gains sufficient robustness to significantly
  improve transferability to previously unseen datasets. In particular, thi
 s enables novel applications for neural rendering where dynamically stream
 ed content is directly incorporated in a (neural) image-based reconstructi
 on of a scene. As we will show, our method reaches state-of-the-art perfor
 mance when compared to previous works that rely on static and sufficiently
  densely sampled scenes; in addition, we demonstrate our system's particul
 ar suitability for dynamically streamed content, where our approach is abl
 e to produce high-fidelity novel-view synthesis even with significantly fe
 wer available frames than competing neural methods.\n\nRegistration Catego
 ry: Full Access, Enhanced Access, Trade Exhibitor, Experience Hall Exhibit
 or\n\n
URL:https://asia.siggraph.org/2023/full-program?id=papers_672&sess=sess209
END:VEVENT
END:VCALENDAR