Warehouse Dataset

Evaluation Criteria for Inside-Out Indoor Positioning Systems based on Machine Learning

Warehouse Dataset

With our indoor logistics Warehouse dataset we aim at providing a solid basis for the development and evaluation of ML-based positioning schemes.

The dataset covers an area of 1,320m² and 464,804 RGB images with a size of 640 x 480 pixels. Each image is labeled with a sub-millimeter position and sub-degree orientation that we acquired using an optical laser-based Nikon iGPS reference system. We recorded the images using a platform with 300mm diameter that carries eight cameras (calibrated Logitech C270) facing in different directions. The distance between the cameras is a few centimeters, which we calibrated out in the labeling.

The dataset includes different scenarios that allow a detailed analysis of positioning schemes based on the evaluation criteria:

Two trajectories through the hall were recorded for training
Eight trajectories were recorded to test the different criteria

Dataset Description

The dataset is split into a training and a testing folder with both containing following sub-folders, where each dataset is available as a downloadable zip file:

	Download	# of pictures	File size
Training
Horizontal	https://www2.iis.fraunhofer.de/IPIN/training/horizontal.tar.gz	98,343	5.9 GB
Vertical	https://www2.iis.fraunhofer.de/IPIN/training/vertical.tar.gz	101,092	5.9 GB
Testing
Cross	https://www2.iis.fraunhofer.de/IPIN/testing/cross.tar.gz	12,375	1.1 GB
Generalize Open	https://www2.iis.fraunhofer.de/IPIN/testing/generalize_open.tar.gz	72,858	4.4 GB
Generalize Racks	https://www2.iis.fraunhofer.de/IPIN/testing/generalize_racks.tar.gz	54,915	3.7 GB
Motion Artifacts	https://www2.iis.fraunhofer.de/IPIN/testing/motion_artifacts_forklift.tar.gz	16,967	1.3 GB
Scale large	https://www2.iis.fraunhofer.de/IPIN/testing/large_scale.tar.gz	43,707	1.0 GB
Scale small	https://www2.iis.fraunhofer.de/IPIN/testing/small_scale.tar.gz	14,458	2.7 GB
Scale Transition	https://www2.iis.fraunhofer.de/IPIN/testing/scale_transition.tar.gz	50,303	3.9 GB
Volatility	https://www2.iis.fraunhofer.de/IPIN/testing/volatility.tar.gz	17,158	1.9 GB
Camera calibration	https://www2.iis.fraunhofer.de/IPIN/camera_calibration.tar.gz

For each camera in each dataset, we provide a text file which stores the path to the images recorded with the according camera and the position and rotation (quaternion) of the camera. The basic scheme of this structure is the following:

path/to/[image].jpg x_pos y_pos z_pos rot_w rot_p rot_q rot_r

We also provide camera calibration matrices for each of the eight cameras. These are found the in the camera calibration zip folder with additional information of the mapping from matrix to cameras.

Licence Agreement

The dataset is released for non-commercial research only. For commercial use, please contact us. If you find it useful, please cite our publication in your work.

Citations

If you report results based on the Warehouse dataset, please cite the paper in your publication. LINK TO PAPER

Löffler, C.; Riechel, S.; Fischer, J.; Mutschler, C.: Evaluation Criteria for Inside-Out Indoor Positioning Systems based on Machine Learning. In: IEEE (Hrsg.): Proceedings of the 9th International Conference on Indoor Positioning and Indoor Navigation (9th International Conference on Indoor Positioning and Indoor Navigation, Nantes, France)., 2018, S. - (BibTeX)