Seminar "Design and evaluation of a biomechanically consistent method for kinematic analysis of markerless sports movements".
The SHV department is pleased to invite you to the seminar by Dr David Pagnon from the University of Bath.
This seminar will take place on Tuesday June 18 from 3:30 to 4:30 pm in room 317 of the Carpeaux building.
-
Le 18/06/2024
-
15:30 - 16:30
-
Seminar
-
Mont Houy Campus
Carpeaux Building
room 317
Title
Design and evaluation of a biomechanically consistent method for kinematic analysis of markerless sports movement
Summary
Motion capture is traditionally carried out using reflective markers. However, these methods are not suitable for contextual analysis of sport on the field, and markerless alternatives are being investigated.
One of the most promising perspectives on this topic lies at the intersection of machine learning for 2D pose estimation, computer vision for 3D reconstruction from multiple video sources, and biomechanics for constraining 3D coordinates to an anatomically consistent model.
We have proposed Pose2Sim, an open-source, easy-to-use package aimed at meeting these needs. OpenPose's 2D detections are robustly triangulated, and passed on to OpenSim for full-body inverse kinematics.
The robustness of Pose2Sim was estimated in the face of "stray" people entering the field of view, degraded image quality, calibration errors, and a reduced number of cameras. Its accuracy was also evaluated, and judged satisfactory for the analysis of walking, running, and cycling.
In a competition context, it may be useful to employ lightweight GoPro-type cameras.
We have tested this equipment on boxing sequences, and proposed post-calibration and post-synchronization procedures.
Finally, capturing both the athlete and his equipment would be interesting. We computed the inverse kinematics of a BMX rider with his bike, training a DeepLabCut model for the bike, triangulated and applied on an OpenSim poly-articulated model.
Finally, capturing both the athlete and his equipment would be interesting.
Taken together, these results provide innovative perspectives for the analysis of sports movement.
Keywords
Motion analysis, Markerless, OpenSim.