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:20230103T035309Z
LOCATION:Room 324\, Level 3\, West Wing
DTSTART;TZID=Asia/Seoul:20221207T140000
DTEND;TZID=Asia/Seoul:20221207T153000
UID:siggraphasia_SIGGRAPH Asia 2022_sess159_papers_156@linklings.com
SUMMARY:MyStyle: A Personalized Generative Prior
DESCRIPTION:Technical Communications, Technical Papers\n\nMyStyle: A Perso
 nalized Generative Prior\n\nNitzan, Aberman, Liba, He, Yarom...\n\nWe intr
 oduce MyStyle, a personalized deep generative prior trained with a few sho
 ts of an individual. MyStyle allows to reconstruct, enhance and edit image
 s of a specific person, such that the output is faithful to the person’s k
 ey facial characteristics. Given a small reference set of portrait images 
 of a person (∼ 100), we tune the weights of a pretrained StyleGAN face gen
 erator to form a local, low-dimensional, personalized manifold in the late
 nt space.\nWe show that this manifold constitutes a personalized region th
 at spans latent codes associated with diverse portrait images of the indiv
 idual. Moreover, we demonstrate that we obtain a personalized generative p
 rior, and propose a unified approach to apply it to various ill-posed imag
 e enhancement problems, such as inpainting and super-resolution, as well a
 s semantic editing. Using the personalized generative prior we obtain outp
 uts that exhibit high-fidelity to the input images and are also faithful t
 o the key facial characteristics of the individual in the reference set. W
 e demonstrate our method with fair-use images of numerous widely recogniza
 ble individuals for whom we have the prior knowledge for a qualitative eva
 luation\nof the expected outcome. We evaluate our approach against few-sho
 ts baselines and show that our personalized prior, quantitatively and qual
 itatively, outperforms state-of-the-art alternatives.\n\nRegistration Cate
 gory: FULL ACCESS, ON-DEMAND ACCESS\n\nLanguage: ENGLISH\n\nFormat: IN-PER
 SON, ON-DEMAND
URL:https://sa2022.siggraph.org/en/full-program/?id=papers_156&sess=sess15
 9
END:VEVENT
END:VCALENDAR