abbott1234 / pyk4a

Python 3 wrapper for Azure-Kinect-Sensor-SDK

pyk4a

This library is a simple and pythonic wrapper in Python 3 for the Azure-Kinect-Sensor-SDK.

Images are returned as numpy arrays and behave like python objects.

This approach incurs almost no overhead in terms of CPU, memory or other resources. It also simplifies usage. Kinect C api image buffers are directly reused and image releases are performed automatically by the python garbage collector.

Prerequisites

The Azure-Kinect-Sensor-SDK is required to build this library. To use the SDK, refer to the installation instructions here.

Install

Linux

Make sure your LD_LIBRARY_PATH contains the directory of k4a.lib

pip install pyk4a

Windows

Make sure you replace the paths in the following instructions with your own k4a sdk path.

pip install pyk4a --global-option=build_ext --global-option="-IC:\Program Files\Azure Kinect SDK v1.2.0\sdk\include" --global-option="-LC:\Program Files\Azure Kinect SDK v1.2.0\sdk\windows-desktop\amd64\release\lib"

Don't forget to add the folder containing the release k4a.dll to your Path env variable C:\Program Files\Azure Kinect SDK v1.2.0\sdk\windows-desktop\amd64\release\bin

Example

For a basic example displaying the first frame, you can run this code:

from pyk4a import PyK4A

# Load camera with the default config
k4a = PyK4A()
k4a.connect()

# Get the next color frame without the depth (blocking function)
img_color = k4a.get_capture(color_only=True)

# Display with pyplot
from matplotlib import pyplot as plt
plt.imshow(img_color[:, :, 2::-1]) # BGRA to RGB
plt.show()

Otherwise, a more avanced example is available in the example folder. To execute it opencv-python is required.

git clone https://github.com/etiennedub/pyk4a.git
cd pyk4a/example
python viewer.py

Documentation

No documentation is available but most methods are used in the example. You can follow it as reference. You can also check directly the code of the main class PyK4A.

Contribution

Feel free to send pull requests.

Bug reports are also appreciated. Please include as much details as possible.

TODO:

• Support for all sensors (other than color and depth)
• get_capture in MJPG