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 B7 (1)\, B Block\, Level 7
DTSTART;TZID=Asia/Tokyo:20241205T105600
DTEND;TZID=Asia/Tokyo:20241205T110800
UID:siggraphasia_SIGGRAPH Asia 2024_sess129_papers_497@linklings.com
SUMMARY:RoMo: A Robust Solver for Full-body Unlabeled Optical Motion Captu
 re
DESCRIPTION:Xiaoyu Pan and Bowen Zheng (State Key Laboratory of CAD&CG, Zh
 ejiang University); Xinwei Jiang, Zijiao Zeng, and Qilong Kou (Tencent Gam
 es Digital Content Technology Center); He Wang (Department of Computer Sci
 ence and UCL Centre for Artificial Intelligence, University College London
 ); and Xiaogang Jin (State Key Laboratory of CAD&CG, Zhejiang University)\
 n\nOptical motion capture (MoCap) is the "gold standard" for accurately ca
 pturing full-body motions. To make use of raw MoCap point data, the system
  labels the points with corresponding body part locations and solves the f
 ull-body motions. However, MoCap data often contains mislabeling, occlusio
 n and positional errors, requiring extensive manual correction. To allevia
 te this burden, we introduce RoMo, an automatic learning-based framework f
 or robustly labeling and solving raw optical motion capture data. In the l
 abeling stage, RoMo employs a divide-and-conquer strategy to break down th
 e complex full-body labeling challenge into manageable subtasks: full-body
  segmentation and part-specific labeling. To utilize the temporal continui
 ty of markers, RoMo generates marker tracklets using a K-partite graph-bas
 ed clustering algorithm, where markers serve as nodes and edges are formed
  based on positional and feature similarities. For motion solving, to prev
 ent error accumulation along the kinematic chain, we introduce a hybrid in
 verse kinematic solver that utilizes joint positions as intermediate repre
 sentations and adjusts the template skeleton to match estimated joint rota
 tions. We demonstrate that RoMo achieves high labeling and solving accurac
 y across multiple metrics and various datasets. Extensive comparisons show
  that our method outperforms state-of-the-art research methods. On a real 
 dataset, RoMo improves the F1 score of hand labeling from 0.94 to 0.98, an
 d reduces the position error of body motion solving by 25%. Furthermore, R
 oMo can be applied in scenarios where commercial systems are inadequate.\n
 \nRegistration Category: Full Access, Full Access Supporter\n\nLanguage Fo
 rmat: English Language\n\nSession Chair: Yuting Ye (Reality Labs Research,
  Meta; Meta)\n\n
URL:https://asia.siggraph.org/2024/program/?id=papers_497&sess=sess129
END:VEVENT
END:VCALENDAR