The Python Kinect Toolkit (PyKinectTk) is a Python package that allows you to extract the data you want from Microsoft extended event files (.xef) generated from the Microsoft Kinect V2, and even from a live stream of data. The package also comes with modules that allow you to play the captured data back, and plot the data on graphs using matplot.
Download the .zip or .tar.gz file and extract its contents. Then, from the command line, change directory to where you extracted the files and run the command python setup.py install and you're good to go! I hope to have this up on pip in the near future.
In the Examples folder there are two files; ex-capture.py and ex-playback.py. These can be used to collect data from a running Kinect Service, and also playback this data to the user.
"""
ex-capture.py
Example script for extracting Kinect Data from Kinect Studio
"""
if __name__ == "__main__":
# Import package
import PyKinectTk
# Create connection to Kinect Service
App = PyKinectTk.Capture.KinectService(timeout=2)
# Start capturing data using auto-click
print "Listening for Kinect data"
App.listen(getVideo=True, Clicking=True)
# Add a meaningful name to the recording
name = raw_input("Would you like to name your recording? ")
App.NameRecording(name)
# Exit
raw_input("Recording saved as '%s', press Return to quit" % name)
App.close()On first import, PyKinectTk checks the contents of PyKinectTk/utils/Settings/config to see if it contains a filepath. If it does not, you will be asked to select a folder to set as your working environment. This creates a number of directories for storing extracted data. You will only be asked to set your working environment on your first use, or if the path to the working environment directory changes. For more info on your working environment, see Your Working Environment.
This is the Python class that talks to the Microsoft Kinect service that is running on the local machine. The timeout argument specifies how long it should wait after receiving a frame of data from the Kinect Service before deciding any data streams have stopped. Once it has been created, you can invoke the the listen() method to begin collecting data that is being processed by the Kinect Service. By default, the Capture.KinectService() only captures skeleton data but you can change this by invoking listen() with keyword "getter" arguments:
App = PyKinectTk.Capture.KinectService(timeout=2)
App.listen(getAudio=True, getVideo=True, getDepth=True, Clicking=True)Note: Depth data capture has not been implemented yet
The Clicking keyword is used to automatically step through files in Kinect Studio, which is useful when you want to process more data intensive streams such as video. Clicking should not be set to True when capturing data from a live stream. For best results when capturing video data, record your data using the Kinect Studio as an .xef file and then process it using the automated clicking method. When doing so, you will be asked to hover over the "step file" button in Kinect Studio and press "Return". This will begin automatically clicking the "step file" button every 0.5 seconds (2fps) to ensure minimal video frames are dropped.
"""
ex-playback.py
Example script for combining extracted
data streams and playing them back to the user
To use the graphical user interface:
python ex-playback.py -GUI
CLI Usage:
python ex-playback.py <recording_id>
python ex-playback.py -n <recording_name>
Other flags:
-video : 1 to display RGB video, 0 to skip. Default is 0.
-info : 1 to display frame no. and time stamp, 0 to skip. Default is 1.
-body : 1 to display wireframes, 0 to skip. Default is 1.
-c <output.avi> : Signals the program to convert the playback data to an .avi file
-t <start:end> : Specifies the timeframe (in seconds) to playback
"""
if __name__ == "__main__":
# Get command arguments from the user
import sys, shlex
args = sys.argv[1:] if sys.argv[1:] else shlex.split(raw_input("Input: "))
if len(args) < 1:
print __doc__
sys.exit()
# Import module
import PyKinectTk
# Create TKinter GUI to select recording
if "-GUI" in args:
application = PyKinectTk.Playback.KinectDataSelect()
sys.exit()
else:
application = PyKinectTk.Playback.KinectDataPlayer
# Can choose a recording by cli
kwargs = {}
# By default, the first argument is the id of the recordings
pid = args[0]
if "-n" in args:
pid = PyKinectTk.Load.PerformanceID(args[args.index("-n") + 1])
if "-c" in args:
application = PyKinectTk.Playback.ConvertKinect
kwargs['outputFile'] = args[args.index("-c") + 1]
if "-t" in args:
time = args[args.index("-t") + 1].split(':')
time = tuple([float(t) if t else None for t in time])
kwargs['time'] = time
if "-video" in args:
kwargs['video'] = bool(args[args.index("-video") + 1])
if "-body" in args:
kwargs['body'] = bool(args[args.index("-body") + 1])
if "-info" in args:
kwargs['info'] = bool(args[args.index("-info") + 1])
# Run application
print "loading..." ,
try:
pid = int(pid)
except:
print "Argument %s is not a valid ID" % repr(pid)
sys.exit()
application(pid, **kwargs).run()
raw_input("done! Press return to exit")Technical documentation for the project is found in the docs.md file found here: PyKinectTk Documentation
This Python package uses a lot of code from PyKinect2, which can be found here: https://github.com/vladkol, and I would like to thank vladkol for his incredible work. His existing code has been adapted and is included as a sub-package as part of this module.