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:20260114T163641Z
LOCATION:Meeting Room C4.8\, Level 4 (Convention Centre)
DTSTART;TZID=Australia/Melbourne:20231214T105000
DTEND;TZID=Australia/Melbourne:20231214T110500
UID:siggraphasia_SIGGRAPH Asia 2023_sess149_papers_210@linklings.com
SUMMARY:Learning the Geodesic Embedding with Graph Neural Networks
DESCRIPTION:Bo Pang (Peking Unversity); Zhongtian Zheng (Peking University
 ); Guoping Wang (Peking Unversity); and Peng-Shuai Wang (Peking University
 , Wangxuan Institute of Computer Technology)\n\nWe present GeGnn, a learni
 ng-based method for computing the approximate geodesic distance between tw
 o arbitrary points on discrete polyhedra surfaces with constant time compl
 exity after fast precomputation. Previous relevant methods either focus on
  computing the geodesic distance between a single source and all destinati
 ons, which has linear complexity at least, or require long precomputation 
 time. Our key idea is to train a graph neural network to embed an input me
 sh into a high-dimensional embedding space and compute the geodesic distan
 ce between a pair of points using the corresponding embedding vectors and 
 a lightweight decoding function. To facilitate the learning of the embeddi
 ng, we propose novel graph convolution and graph pooling modules that inco
 rporate local geodesic information and are verified to be much more effect
 ive than previous designs. After training, our method requires only one fo
 rward pass of the network per mesh as precomputation. Then, we can compute
  the geodesic distance between a pair of points using our decoding functio
 n, which requires only several matrix multiplications and can be massively
  parallelized on GPUs. We verify the efficiency and effectiveness of our m
 ethod on ShapeNet and demonstrate that our method is faster than existing 
 methods by orders of magnitude while achieving comparable or better accura
 cy. Additionally, our method exhibits robustness on noisy and incomplete m
 eshes and strong generalization ability on out-of-distribution meshes.\n\n
 Registration Category: Full Access\n\nSession Chair: Marco ATTENE (Institu
 te for Applied Mathematics and Information Technologies (IMATI), CNR)\n\n
URL:https://asia.siggraph.org/2023/full-program?id=papers_210&sess=sess149
END:VEVENT
END:VCALENDAR