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:20230103T035312Z
LOCATION:Room 324\, Level 3\, West Wing
DTSTART;TZID=Asia/Seoul:20221209T090000
DTEND;TZID=Asia/Seoul:20221209T103000
UID:siggraphasia_SIGGRAPH Asia 2022_sess172_papers_112@linklings.com
SUMMARY:Efficient Neural Radiance Fields for Interactive Free-viewpoint Vi
 deo
DESCRIPTION:Technical Communications, Technical Papers\n\nEfficient Neural
  Radiance Fields for Interactive Free-viewpoint Video\n\nLin, Peng, Xu, Ya
 n, Shuai...\n\nThis paper aims to tackle the challenge of efficiently prod
 ucing interactive free-viewpoint videos. \nSome recent works equip neural 
 radiance fields with image encoders, enabling them to generalize across sc
 enes. When processing dynamic scenes, they can simply treat each video fra
 me as an individual scene and perform novel view synthesis to generate fre
 e-viewpoint videos. However, their rendering process is slow and cannot su
 pport interactive applications. \nA major factor is that they sample lots 
 of points in empty space when inferring radiance fields. \nWe propose a no
 vel scene representation, called ENeRF, for the fast creation of interacti
 ve free-viewpoint videos. Specifically, given multi-view images at one fra
 me, we first build the cascade cost volume to predict the coarse geometry 
 of the scene. The coarse geometry allows us to sample few points near the 
 scene surface, thereby significantly improving the rendering speed. This p
 rocess is fully differentiable, enabling us to jointly learn the depth pre
 diction and radiance field networks from RGB images. Experiments show that
  our approach exhibits competitive performance on the DTU, NeRF Synthetic,
  Real Forward-facing, ZJU-MoCap, and DynamicCap datasets while being at le
 ast 60 times faster than previous generalizable radiance field methods. We
  demonstrate the capability of our method to synthesize novel views of hum
 an performers in real-time.  The code is available at https://zju3dv.githu
 b.io/enerf/.\n\nRegistration Category: FULL ACCESS, ON-DEMAND ACCESS\n\nLa
 nguage: ENGLISH\n\nFormat: IN-PERSON, ON-DEMAND
URL:https://sa2022.siggraph.org/en/full-program/?id=papers_112&sess=sess17
 2
END:VEVENT
END:VCALENDAR