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:20260817T171532Z
LOCATION:Hall B5 (1)\, B Block\, Level 5
DTSTART;TZID=Asia/Tokyo:20241206T135600
DTEND;TZID=Asia/Tokyo:20241206T141000
UID:siggraphasia_SIGGRAPH Asia 2024_sess145_papers_1060@linklings.com
SUMMARY:Multi-level Partition of Unity on Differentiable Moving Particles
DESCRIPTION:Jinjin He and Taiyuan Zhang (Dartmouth College); Hiroki Kobaya
 shi and Atsushi Kawamoto (Toyota Central R&D Labs., Inc.); Yuqing Zhou (To
 yota Research Institute of North America); Tsuyoshi Nomura (Toyota Central
  R&D Labs., Inc.); and Bo Zhu (Georgia Institute of Technology)\n\nWe intr
 oduce a differentiable moving particle representation based on the\nmulti-
 level partition of unity (MPU) to represent dynamic implicit geome-\ntries
 . At the core of our representation are two groups of particles, named\nfe
 ature particles and sample particles, which can move in space and produce\
 ndynamic surfaces according to external velocity fields or optimization gr
 adi-\nents. These two particle groups iteratively guide and correct each o
 ther by\nalternating their roles as inputs and outputs. Each feature parti
 cle carries\na set of coefficients for a local quadratic patch. These part
 icle patches are\nassembled with partition-of-unity weights to derive a co
 ntinuous implicit\nglobal shape. Each sampling particle carries its positi
 on and orientation,\nserving as dense surface samples for optimization tas
 ks. Based on these mov-\ning particles, we develop a fully differentiable 
 framework to infer and evolve\nhighly detailed implicit geometries, enhanc
 ed by a multi-level background\ngrid for particle adaptivity, across diffe
 rent inverse tasks. We demonstrated\nthe efficacy of our representation th
 rough various benchmark comparisons\nwith state-of-the-art neural represen
 tations, achieving lower memory con-\nsumption, fewer training iterations,
  and orders of magnitude higher accuracy\nin handling topologically comple
 x objects and dynamic tracking tasks.\n\nRegistration Category: Full Acces
 s, Full Access Supporter\n\nLanguage Format: English Language\n\nSession C
 hair: Hao (Richard) Zhang (Simon Fraser University, Augmenta)\n\n
URL:https://asia.siggraph.org/2024/program/?id=papers_1060&sess=sess145
END:VEVENT
END:VCALENDAR