Tools for constructing 3D pose tracks from multi-camera 2D poses
To install both the Python library and the command line interface, launch a
Python virtual environment (optional, but recommended) and then install via
pip
:
pip install poseconnect
The command line interface consists of a single command (poseconnect
) with
four subcommands:
reconstruct
takes a file of 2D pose data and a file of camera calibration info as inputs and outputs a file of 3D pose datatrack
takes a file of 3D pose data as input and outputs a file of 3D pose data with track IDsinterpolate
takes a file of 3D pose data with track IDs as input and fills time gaps in each track with linearly interpolated posesidentify
takes a file of 3D pose data with track IDs and a file of location sensor data with person IDs and outputs a file of 3D pose data with track and person IDsoverlay
takes a file of 2D or 3D pose data and a source video for the data and outputs a video with the poses overlaid
Overall help is available at poseconnect --help
and help for each subcommand
(including usage and a list of options) is available at poseconnect SUBCOMMAND --help
(e.g., poseconnect reconstruct --help
).
More information about input and output data file formats is available here.
To see the command line interface in action, download and unzip the following sample JSON data files to your current directory and run the demo shell script: