This repository provides a demo for motion tracking from multiple cameras using OpenPose, and subsequent triangulation of 2D coordinates using pose2sim. The 3D coordinates can be used, for instance, in OpenSim that allows for extraction of full-body 3D joint angles.
This workflow demo is quite experimental and work in progress, but because it has all that you need to test it, we believe it can be helpful and is a little bit quicker to set up than the original documentation of pose2sim. You can easily adapt it to your own workflow.
Please study pose2sim package that we are using here: https://github.com/perfanalytics/pose2sim?tab=readme-ov-file#installation.
-- Images = Some images and temporary video folder
-- openpose = OpenPose materials, see inside (You need to seperately download bin and 25B models)
-- Pose2Sim = pose2sim projects files
-- projectdata = each folder within projectdata contains a trial or seperate scene, containing calibration folder, and raw 2D videos (then folders are added as you run the code)
-- User = some logging that pose2sim does for archiving output during running of code
openpose_to_pose2sim_tracking.ipynb = main jupyter notebook script to run the envision demo
conda create --name 3Dtracking python=3.8
conda activate 3Dtracking
cd The/path/of/the/downloaded/folder
pip install -r requirements.txt
python -m ipykernel install --user --name=envname
citation for this module: Kadavá, S., Pouw, W. (2024). 3D motion tracking using multiple cameras with OpenPose and Pose2Sim [the day you viewed the site]. Retrieved from: xxx