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:20250110T023312Z
LOCATION:Hall B5 (2)\, B Block\, Level 5
DTSTART;TZID=Asia/Tokyo:20241204T133400
DTEND;TZID=Asia/Tokyo:20241204T134600
UID:siggraphasia_SIGGRAPH Asia 2024_sess116_papers_480@linklings.com
SUMMARY:TurboEdit: Text-Based Image Editing Using Few-Step Diffusion Model
 s
DESCRIPTION:Technical Papers\n\nGilad Deutch (Tel Aviv University); Rinon 
 Gal (Tel Aviv University, NVIDIA Research); and Daniel Garibi, Or Patashni
 k, and Daniel Cohen-Or (Tel Aviv University)\n\nDiffusion models have open
 ed the path to a wide range of text-based image editing frameworks. Howeve
 r, these typically build on the multi-step nature of the diffusion backwar
 ds process, and adapting them to distilled, fast-sampling methods has prov
 en surprisingly challenging. Here, we focus on a popular line of text-base
 d editing frameworks - the "edit-friendly" DDPM-noise inversion approach. 
 We analyze its application to fast sampling methods and categorize its fai
 lures into two classes: the appearance of visual artifacts, and insufficie
 nt editing strength. We trace the artifacts to mismatched noise statistics
  between inverted noises and the expected noise schedule, and suggest a sh
 ifted noise schedule which corrects for this offset. To increase editing s
 trength, we propose a pseudo-guidance approach that efficiently increases 
 the magnitude of edits without introducing new artifacts. All in all, our 
 method enables text-based image editing with as few as three diffusion ste
 ps, while providing novel insights into the mechanisms behind popular text
 -based editing approaches.\n\nRegistration Category: Full Access, Full Acc
 ess Supporter\n\nLanguage Format: English Language\n\nSession Chair: Dani 
 Lischinski (Hebrew University of Jerusalem, Google)
URL:https://asia.siggraph.org/2024/program/?id=papers_480&sess=sess116
END:VEVENT
END:VCALENDAR