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:20240214T070241Z
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_430@linklings.com
SUMMARY:Domain-Agnostic Tuning-Encoder for Fast Personalization of Text-To
 -Image Models
DESCRIPTION:Technical Papers\n\nMoab Arar (Tel-Aviv University); Rinon Gal
  (Tel Aviv University, NVIDIA Research); Yuval Atzmon (NVIDIA Research); G
 al Chechik (NVIDIA Research, Bar-Ilan University); Daniel Cohen-Or (Tel Av
 iv University); Ariel Shamir (Reichman University (IDC)); and Amit H. Berm
 ano (Tel Aviv University)\n\nText-to-image (T2I) personalization allows us
 ers to guide the creative image generation process by combining their own 
 visual concepts in natural language prompts. \nRecently, encoder-based tec
 hniques have emerged as a new effective approach for T2I personalization, 
 reducing the need for multiple images and long training times.\nHowever, m
 ost existing encoders are limited to a single-class domain, which hinders 
 their ability to handle diverse concepts. In this work, we propose a domai
 n-agnostic method that does not require any specialized dataset or prior i
 nformation about the personalized concepts. We introduce a novel contrasti
 ve-based regularization technique to maintain high fidelity to the target 
 concept characteristics while keeping the predicted embeddings close to ed
 itable regions of the latent space, by pushing the predicted tokens toward
  their nearest existing CLIP tokens. Our experimental results demonstrate 
 the effectiveness of our approach and show how the learned tokens are more
  semantic than tokens predicted by unregularized models. This leads to a b
 etter representation that achieves state-of-the-art performance while bein
 g more flexible than previous methods.\n\nRegistration Category: Full Acce
 ss, Enhanced Access, Trade Exhibitor, Experience Hall Exhibitor
URL:https://asia.siggraph.org/2023/full-program?id=papers_430&sess=sess209
END:VEVENT
END:VCALENDAR