BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Asia/Seoul
X-LIC-LOCATION:Asia/Seoul
BEGIN:STANDARD
TZOFFSETFROM:+0900
TZOFFSETTO:+0900
TZNAME:KST
DTSTART:18871231T000000
DTSTART:19881009T020000
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20230103T035307Z
LOCATION:Room 325-AB\, Level 3\, West Wing
DTSTART;TZID=Asia/Seoul:20221206T140000
DTEND;TZID=Asia/Seoul:20221206T153000
UID:siggraphasia_SIGGRAPH Asia 2022_sess156_papers_167@linklings.com
SUMMARY:SHRED: 3D Shape Region Decomposition with Learned Local Operations
DESCRIPTION:Technical Communications, Technical Papers, TOG\n\nSHRED: 3D S
 hape Region Decomposition with Learned Local Operations\n\nJones, Habib, R
 itchie\n\nWe present SHRED, a method for 3D SHape REgion Decomposition. SH
 RED takes a 3D point cloud as input and uses learned local operations to p
 roduce a segmentation that approximates fine-grained part instances. We en
 dow SHRED with three decomposition operations: splitting regions, fixing t
 he boundaries between regions, and merging regions together. Modules are t
 rained independently and locally, allowing SHRED to generate high-quality 
 segmentations for categories not seen during training. We train and evalua
 te SHRED with fine-grained segmentations from PartNet; using its merge-thr
 eshold hyperparameter, we show that SHRED produces segmentations that bett
 er respect ground-truth annotations compared with baseline methods, at any
  desired decomposition granularity. Finally, we demonstrate that SHRED is 
 useful for downstream applications, out-performing all baselines on zero-s
 hot fine-grained part instance segmentation and few-shot fine-grained sema
 ntic segmentation when combined with methods that learn to label shape reg
 ions.\n\nRegistration Category: FULL ACCESS, ON-DEMAND ACCESS\n\nLanguage:
  ENGLISH\n\nFormat: IN-PERSON, ON-DEMAND
URL:https://sa2022.siggraph.org/en/full-program/?id=papers_167&sess=sess15
 6
END:VEVENT
END:VCALENDAR