Multimedia Computing and Computer Vision Lab












Student Theses


Source Code / Datasets





From Multimedia Computing Lab - University of Augsburg

KAET: Kinect Annotation and Evaluation Tool

The Kinect Annotation and Evaluation Tool is made to simplify the process of creating and improving machine learning algorithms for 2D and 3D pose estimation. It offers an intuitive user interface to do some of the most common tasks in this context: get annotated images (optionally background separated) containing one or more persons and train and evaluate a machine learning algorithm with these annotations.

Recorded material can be browsed with the graphical user interface and frames can be exported to single image files. A simple annotation xml file format makes it easy to train algorithms with this data. Finally, it is possible to find and analyse possible flaws of algorithms with an interactive error chart.

The project and the source code is available freely under the terms of the GPL v3. A detailed documentation comes with the software and is available in the application folder.


  • .NET Framework v4.0 or higher,
  • A computer with 2Gb ram and 2.6 Ghz dual core processor or better,
  • OpenNI (installation instructions are included in the README document),
  • Only for recording: Kinect, NITE and SensorKinect driver (installation instructions are included in the README document), sufficient free hard disk space and a fast hard drive.


For more information please contact Christoph Lassner or Thomas Greif.