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:20240214T070247Z
LOCATION:Meeting Room C4.8\, Level 4 (Convention Centre)
DTSTART;TZID=Australia/Melbourne:20231214T110500
DTEND;TZID=Australia/Melbourne:20231214T111500
UID:siggraphasia_SIGGRAPH Asia 2023_sess149_papers_778@linklings.com
SUMMARY:Zero-Shot 3D Shape Correspondence
DESCRIPTION:Technical Papers\n\nAhmed Abdelreheem and Abdelrahman Eldesoke
 y (King Abdullah University of Science and Technology (KAUST)), Maks Ovsja
 nikov (Centre National de la Recherche Scientifique - Laboratoire d'inform
 atique de l'École Polytechnique (LIX)), and Peter Wonka (King Abdullah Uni
 versity of Science and Technology (KAUST))\n\nWe propose a novel zero-shot
  approach to computing correspondences\nbetween 3D shapes. Existing approa
 ches mainly focus on isometric and\nnear-isometric shape pairs (e.g., huma
 n vs. human), but less attention has\nbeen given to strongly non-isometric
  and inter-class shape matching (e.g., human vs. cow). To this end, we int
 roduce a fully automatic method that exploits the exceptional reasoning ca
 pabilities of recent foundation models in language and vision to tackle di
 fficult shape correspondence problems. Our approach comprises multiple sta
 ges. First, we classify the 3D shapes in a zero-shot manner by feeding ren
 dered shape views to a language-vision model (e.g., BLIP2) to generate a l
 ist of class proposals per shape. These proposals are unified into a singl
 e class per shape by employing the reasoning capabilities of ChatGPT. Seco
 nd, we attempt to segment the two shapes in a zero-shot manner, but in con
 trast to the co-segmentation problem, we do not require a mutual set of se
 mantic regions. Instead, we propose to exploit the in-context learning cap
 abilities of ChatGPT to generate two different sets of semantic regions fo
 r each shape and a semantic mapping between them. This enables our approac
 h to match strongly non-isometric shapes with significant differences in g
 eometric structure. Finally, we employ the generated semantic mapping to p
 roduce coarse correspondences that can further be refined by the functiona
 l maps framework to produce dense point-to-point maps. Our approach, despi
 te its simplicity, produces highly plausible results in a zero-shot manner
 , especially between strongly non-isometric shapes.\n\nRegistration Catego
 ry: Full Access\n\nSession Chair: Marco ATTENE (Institute for Applied Math
 ematics and Information Technologies (IMATI), CNR)
URL:https://asia.siggraph.org/2023/full-program?id=papers_778&sess=sess149
END:VEVENT
END:VCALENDAR