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:20250110T023309Z
LOCATION:Hall B7 (1)\, B Block\, Level 7
DTSTART;TZID=Asia/Tokyo:20241203T145600
DTEND;TZID=Asia/Tokyo:20241203T150800
UID:siggraphasia_SIGGRAPH Asia 2024_sess108_papers_214@linklings.com
SUMMARY:SD-𝜋XL: Generating Low-Resolution Quantized Imagery via Score Dist
 illation
DESCRIPTION:Technical Papers\n\nAlexandre Binninger and Olga Sorkine-Hornu
 ng (ETH Zürich)\n\nLow-resolution quantized imagery, such as pixel art, is
  seeing a revival in modern applications ranging from video game graphics 
 to digital design and fabrication, where creativity is often bound by a li
 mited palette of elemental units. Despite their growing popularity, the au
 tomated generation of quantized images from raw inputs remains a significa
 nt challenge, often necessitating intensive manual input. We introduce SD-
 𝜋XL, an approach for producing quantized images that employs score distill
 ation sampling in conjunction with a differentiable image generator. Our m
 ethod enables users to input a prompt and optionally an image for spatial 
 conditioning, set any desired output size 𝐻×𝑊 , and choose a palette of 𝑛 
 colors or elements. Each color corresponds to a distinct class for our gen
 erator, which operates on an 𝐻×𝑊×𝑛 tensor. We adopt a softmax approach, co
 mputing a convex sum of elements, thus rendering the process differentiabl
 e and amenable to backpropagation. We show that employing Gumbel-softmax r
 eparameterization allows for crisp pixel art effects. Unique to our method
  is the ability to transform input images into low-resolution, quantized v
 ersions while retaining their key semantic features. Our experiments valid
 ate SD-𝜋XL’s performance in creating visually pleasing and faithful repres
 entations, consistently outperforming the current state-of-the-art. Furthe
 rmore, we showcase SD-𝜋XL’s practical utility in fabrication through its a
 pplications in interlocking brick mosaic, beading and embroidery design.\n
 \nRegistration Category: Full Access, Full Access Supporter\n\nLanguage Fo
 rmat: English Language\n\nSession Chair: I-Chao Shen (The University of To
 kyo)
URL:https://asia.siggraph.org/2024/program/?id=papers_214&sess=sess108
END:VEVENT
END:VCALENDAR
