Textile and Apparel Science

Leveraging Pose Estimation for Markerless Motion Analysis in Naturalistic Settings

---

Abstract

Pose estimation enables markerless motion analysis from ordinary videos, offering a low-cost alternative to laboratory motion capture for studying movement in naturalistic settings. This proceeding presents a pilot workflow applying the Ultralytics YOLOv8x-pose model to publicly available footage of an oil-and-gas derrickman performing high-mobility tasks. Key frames were sampled to capture distinct postural moments (approach, exertion, and load handling). Seventeen 2D keypoints were extracted and used to compute apparel-relevant joint angles (shoulder, elbow, trunk, hip, and knee) via vector geometry. Beyond single-frame snapshots, time-varying knee and elbow profiles were examined across pose transitions to identify when movement demands peak and when garments may constrain mobility. Findings demonstrate the feasibility of field-based, markerless assessment for protective and performance clothing, while highlighting limitations of 2D depth ambiguity and viewpoint-dependent angle estimates. Future work will integrate 3D pose estimation and avatar-based pipelines for virtual prototyping to connect motion data with pattern decisions.

Keywords: Pose estimation, markerless motion capture, joint-angle analysis, occupational movement, protective clothing

How to Cite:

Jung, U., (2025) “Leveraging Pose Estimation for Markerless Motion Analysis in Naturalistic Settings”, International Textile and Apparel Association Annual Conference Proceedings 82(1). doi: https://doi.org/10.31274/itaa.22055