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:20230103T035306Z
LOCATION:Auditorium\, Level 5\, West Wing
DTSTART;TZID=Asia/Seoul:20221206T100000
DTEND;TZID=Asia/Seoul:20221206T120000
UID:siggraphasia_SIGGRAPH Asia 2022_sess153_papers_487@linklings.com
SUMMARY:DynaGAN: Dynamic Few-shot Adaptation of GANs to Multiple Domains
DESCRIPTION:Technical Papers\n\nDynaGAN: Dynamic Few-shot Adaptation of GA
 Ns to Multiple Domains\n\nKim, Kang, Kim, Baek, Cho\n\nFew-shot domain ada
 ptation to multiple domains aims to learn a complex image distribution acr
 oss multiple domains from a few training images. A naive solution here is 
 to train a separate model for each domain using few-shot domain adaptation
  methods. Unfortunately, this approach mandates linearly-scaled computatio
 nal resources both in memory and computation time and, more importantly, s
 uch separate models cannot exploit the shared knowledge between target dom
 ains. In this paper, we propose DynaGAN, a novel few-shot domain-adaptatio
 n method for multiple target domains. DynaGAN has an adaptation module, wh
 ich is a hyper-network that dynamically adapts a pretrained GAN model into
  the multiple target domains. Hence, we can fully exploit the shared knowl
 edge across target domains and avoid the linearly-scaled computational req
 uirements. As it is still computationally challenging to adapt a large-siz
 e GAN model, we design our adaptation module to be lightweight using the r
 ank-1 tensor decomposition. Lastly, we propose a contrastive-adaptation lo
 ss suitable for multi-domain few-shot adaptation. We validate the effectiv
 eness of our method through extensive qualitative and quantitative evaluat
 ions.\n\nRegistration Category: FULL ACCESS, EXPERIENCE PLUS ACCESS, EXPER
 IENCE ACCESS, TRADE EXHIBITOR\n\nLanguage: ENGLISH\n\nFormat: IN-PERSON
URL:https://sa2022.siggraph.org/en/full-program/?id=papers_487&sess=sess15
 3
END:VEVENT
END:VCALENDAR