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:20241206T150800
DTEND;TZID=Asia/Tokyo:20241206T151900
UID:siggraphasia_SIGGRAPH Asia 2024_sess150_papers_189@linklings.com
SUMMARY:World-Grounded Human Motion Recovery via Gravity-View Coordinates
DESCRIPTION:Zehong Shen, Huaijin Pi, Yan Xia, Zhi Cen, and Sida Peng (Stat
 e Key Laboratory of CAD&CG, Zhejiang University); Zechen Hu (Deep Glint); 
 Hujun Bao (State Key Laboratory of CAD&CG, Zhejiang University); Ruizhen H
 u (Shenzhen University (SZU)); and Xiaowei Zhou (State Key Laboratory of C
 AD&CG, Zhejiang University)\n\nWe present a novel method for recovering wo
 rld-grounded human motion from monocular video. The main challenge lies in
  the ambiguity of defining the world coordinate system, which varies betwe
 en sequences. Previous approaches attempt to alleviate this issue by predi
 cting relative motion in an autoregressive manner, but are prone to accumu
 lating errors. Instead, we propose estimating human poses in a novel Gravi
 ty-View (GV) coordinate system, which is defined by the world gravity and 
 the camera view direction. The proposed GV system is naturally gravity-ali
 gned and uniquely defined for each video frame, largely reducing the ambig
 uity of learning image-pose mapping. The estimated poses can be transforme
 d back to the world coordinate system using camera rotations, forming a gl
 obal motion sequence. Additionally, the per-frame estimation avoids error 
 accumulation in the autoregressive methods. Experiments on in-the-wild ben
 chmarks demonstrate that our method recovers more realistic motion in both
  the camera space and world-grounded settings, outperforming state-of-the-
 art methods in both accuracy and speed. The code is available at https://z
 ju3dv.github.io/gvhmr.\n\nRegistration Category: Full Access, Full Access 
 Supporter\n\nLanguage Format: English Language\n\nSession Chair: Li-Yi Wei
  (Adobe Research)\n\n
URL:https://asia.siggraph.org/2024/program/?id=papers_189&sess=sess150
END:VEVENT
END:VCALENDAR