BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/Chicago
X-LIC-LOCATION:America/Chicago
BEGIN:DAYLIGHT
TZOFFSETFROM:-0600
TZOFFSETTO:-0500
TZNAME:CDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0500
TZOFFSETTO:-0600
TZNAME:CST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20251016T135127Z
LOCATION:Grand C/D South
DTSTART;TZID=America/Chicago:20251014T135000
DTEND;TZID=America/Chicago:20251014T141000
UID:HFESAM_ASPIRE 2025_sess237_LECT327@linklings.com
SUMMARY:Designing and Building a 3D Human Motion Dataset for Vision-Based 
 Ergonomics Risk Assessments
DESCRIPTION:Leyang Wen (University of Michigan), Daeho Kim (University of 
 Toronto), Veeru Talreja (VelocityEHS), Meiyin Liu (Rutgers University), Ju
 lia Penfield and Rick Barker (VelocityEHS), and SangHyun Lee (University o
 f Michigan)\n\nArtificial Intelligence (AI) is increasingly used in ergono
 mics, particularly for assessing musculoskeletal disorder (MSD) risks usin
 g computer vision-based evaluation. Recent advancements in vision-based AI
  have made it possible to monitor MSD risks using an ordinary camera, offe
 ring a more accessible and less intrusive alternative to traditional obser
 vation methods. However, existing AI models, often trained on generic data
 sets from the computer science domain, lack the keypoints necessary for ca
 lculating the intricate angles, especially in high-degree-of-freedom (DoF)
  joints, like distinguishing neck flexion/extension, lateral flexion, and 
 rotation, for ergonomics risk assessments. We present the design and build
 ing process of a large-scale 3D human motion dataset designed to train vis
 ion-based AI models for ergonomics risk assessments. The dataset includes 
 47 custom-selected keypoints optimized for high-DoF joint angle calculatio
 ns and visual features, capturing 7 million frames of 10 subjects performi
 ng 9 categories of manual material handling tasks under varied task parame
 ters. A baseline MotionBert model trained on our dataset achieved a mean a
 bsolute angle error of 3.5° on the validation set and demonstrated general
 ization capability on real-world industry videos. Our work offers guidelin
 es for collecting custom datasets and training AI models to enable more ac
 cessible, less intrusive monitoring of MSD risks using ordinary cameras.\n
 \nTrack: Occupational Ergonomics\n\nSession Chairs: Saman Madinei (Meta) a
 nd Mina Salehi (Oregon State University)\n\n
END:VEVENT
END:VCALENDAR