BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Asia/Tokyo
X-LIC-LOCATION:Asia/Tokyo
BEGIN:STANDARD
TZOFFSETFROM:+0900
TZOFFSETTO:+0900
TZNAME:JST
DTSTART:18871231T000000
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20260817T171530Z
LOCATION:Hall B7 (1)\, B Block\, Level 7
DTSTART;TZID=Asia/Tokyo:20241203T133400
DTEND;TZID=Asia/Tokyo:20241203T134600
UID:siggraphasia_SIGGRAPH Asia 2024_sess105_papers_697@linklings.com
SUMMARY:Customizing Text-to-Image Models with a Single Image Pair
DESCRIPTION:Maxwell Jones, Sheng-Yu Wang, and Nupur Kumari (Carnegie Mello
 n University); David Bau (Northeastern University); and Jun-Yan Zhu (Carne
 gie Mellon University)\n\nArt reinterpretation is the practice of creating
  a variation of a reference work, making a paired artwork that exhibits a 
 distinct artistic style. We ask if such an image pair can be used to custo
 mize a generative model to capture the demonstrated stylistic difference. 
 We propose Pair Customization, a new customization method that learns styl
 istic difference from a single image pair and then applies the acquired st
 yle to the generation process. Unlike existing methods that learn to mimic
  a single concept from a collection of images, our method captures the sty
 listic difference between paired images. This allows us to apply a stylist
 ic change without overfitting to the specific image content in the example
 s. To address this new task, we employ a joint optimization method that ex
 plicitly separates the style and content into distinct LoRA weight spaces.
  We optimize these style and content weights to reproduce the style and co
 ntent images while encouraging their orthogonality. During inference, we m
 odify the diffusion process via a new style guidance based on our learned 
 weights. Both qualitative and quantitative experiments show that our metho
 d can effectively learn style while avoiding overfitting to image content,
  highlighting the potential of modeling such stylistic differences from a 
 single image pair.\n\nRegistration Category: Full Access, Full Access Supp
 orter\n\nLanguage Format: English Language\n\nSession Chair: Kfir Aberman 
 (Decart AI)\n\n
URL:https://asia.siggraph.org/2024/program/?id=papers_697&sess=sess105
END:VEVENT
END:VCALENDAR