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:20230103T035308Z
LOCATION:Room 324\, Level 3\, West Wing
DTSTART;TZID=Asia/Seoul:20221207T110000
DTEND;TZID=Asia/Seoul:20221207T123000
UID:siggraphasia_SIGGRAPH Asia 2022_sess158_papers_421@linklings.com
SUMMARY:Reconstructing Personalized Semantic Facial NeRF Models From Monoc
ular Video
DESCRIPTION:Technical Papers\n\nReconstructing Personalized Semantic Facia
l NeRF Models From Monocular Video\n\nGao, Zhong, Xiang, Hong, Guo...\n\nW
e present a novel semantic model for human head defined with neural radian
ce field. The 3D-consistent head model consist of a set of disentangled an
d interpretable bases, and can be driven by low-dimensional expression coe
fficients. Thanks to the powerful representation ability of neural radianc
e field, the constructed model can represent complex facial attributes inc
luding hair, wearings, which can not be represented by traditional mesh bl
endshape. To construct the personalized semantic facial model, we propose
to define the bases as several multi-level voxel fields. With a short mono
cular RGB video as input, our method can construct the subject's semantic
facial NeRF model with only ten to twenty minutes, and can render a photo-
realistic human head image in tens of miliseconds with a given expression
coefficient and view direction. With this novel representation, we apply i
t to many tasks like facial retargeting and expression editing. Experiment
al results demonstrate its strong representation ability and training/infe
rence speed.\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_421&sess=sess15
8
END:VEVENT
END:VCALENDAR