Cognitive navigation dataset

The creation of this dataset is motivated by the wish to provide to the research community a handy tool for the validation of localization and place classification algorithms. It has been captured by exploiting an RGB-D sensor, i.e. a Microsoft Kinect, mounted on the MAGGIE (Mobile Autonomous riGGed Indoors Exploratory) Robot. The Cognitive Navigation Dataset consists a long range indoors route, which has been recorded in the premises of the Democritus University of Thrace to serve the needs of the Group of Robotics and Cognitive Systems.

I would like to thank my colleagues Mr. Evangelos Boukas and Mr. Dimitris Chrysostomou for their assistance during the acquisition of this dataset. 

The dataset contains color images accompanied by the respective depth maps of the scene as they have been obtained directly from a Microsoft Kinect sensor. It consisted from three different parts:

  1. The first part (Part A) involves the representation of a "corridor", an "office", a "laboratory", an "elevator area" and a "bathroom" in natural illumination conditions. This means that the first part has been captured using the available day light.

The second part (Part B) involves exactly the same places as the Part A but it has been captured at night and, therefore, the illumination conditions are artificial.

•The third part (Part C) of the dataset has also been acquired under artificial illumination conditions and comprises a continuous exploration of all the aforementioned room places. The movements of the robot are very smooth with common field of view between the successive frames.

The first two parts of the dataset are suitable for the validation of place classification algorithms i.e. recognition and categorization due to the variation in the illumination conditions.  The third part of the dataset is suitable for indoor localization and mapping due to the fact that the consecutive frames have great spatial and time proximity. Additionally, the color images in all the three parts are accompanied by their respective depth images.  This Cognitive Navigation Dataset is freely available to the research community for download. More technical information about the robot set-up and the download link are available here.

An outline of the various scenes of Cognitive Navigation Dataset is presented below:

Part A.  Data images captured with natural illumination conditions

Part B.  Data images captured with artificial illumination conditions

Part C. Localization images

Download link and further technical information