Datasets from the ATC shopping center
As part of our project on enabling mobile social robots to work in public spaces (project homepage in Japanese, founded by JST/CREST) we set up a tracking environment in the "ATC" shopping center in Osaka, Japan. The system consists of multiple 3D range sensors, covering an area of about 900 m2. You can find more details about the system in the reference paper below.
Here is a video showing the tracking system in action:
We provide the complete person tracking dataset as well as a sample of the raw sensor measurements.
ATC pedestrian tracking dataset
The data provided here was collected between October 24, 2012 and November 29, 2013. In general the data collection was done every week on Wednesday and Sunday, from morning until evening (9:40-20:20). The dataset consists of 92 days in total. Refer to the list of days for details on the included dates.
The data is provided as CSV files, one file for each day (file names are in the format atc-YYYYMMDD.csv). Each row in a CSV file corresponds to a single tracked person at a single instant, and it contains the following fields:
time [ms] (unixtime + milliseconds/1000), person id, position x [mm], position y [mm], position z (height) [mm], velocity [mm/s], angle of motion [rad], facing angle [rad]
The dataset is split into 9 parts, each consisting of 10 or 11 single-day CSV files compressed together in a 7z archive. The files are big, so downloading might take a while. (If you just want to try the data out, here is a single day sample [201 MB]).
Tracking data - part 1 [3.9 GB]
Tracking data - part 2 [4.2 GB]
Tracking data - part 3 [3.9 GB]
Tracking data - part 4 [4.4 GB]
Tracking data - part 5 [4.4 GB]
Tracking data - part 6 [4.7 GB]
Tracking data - part 7 [4.1 GB]
Tracking data - part 8 [5.1 GB]
Tracking data - part 9 [4.6 GB]
You can also download a sample SQL script for loading the uncompressed CSV files into an SQL database.
Raw sensor data sample
This is a sample of the raw measurement data taken from all 49 3D range sensors installed in ATC. The data is for the period from 13:00 to 13:15 on November 14, 2012. (A larger dataset can be provided upon request.)
The files are compressed as tar.gz. They are rather big, so downloading might take a while.
Raw data 1 (sensors A1-11, V1-2) [2.9 GB]
Raw data 2 (sensors P1-12) [4.0 GB]
Raw data 3 (sensors P13-24) [3.8 GB]
Raw data 4 (sensors P24-36) [3.9 GB]
Sensor description and calibration details (xlsx)
Sample data reading program (Java)
Sample data reading program with conversion to 3d points (Python, added May 2023)
Extra files
2D occupancy grid of the area, which was built using a robot equipped with a 3D laser scanner. The files (localization_grid.pgm and localization_grid.yaml) are in the format used by map_server in ROS. The map uses the same coordinate system as the ATC dataset so it can be aligned with the pedestrian data.
Slide (.pptx format) containing the floor plan with extra details about the tracking data and the tracking area.
License
The datasets are free to use for research purposes only. In case you use the datasets in your work please be sure to cite the reference paper below.
Reference
D. Brscic, T. Kanda, T. Ikeda, T. Miyashita, "Person position and body direction tracking in large public spaces using 3D range sensors", IEEE Transactions on Human-Machine Systems, Vol. 43, No. 6, pp. 522-534, 2013
Related papers
You might also like to check out the works where we analyzed the tracking data:
- D. Brscic, T. Kanda, "Changes in Usage of an Indoor Public Space: Analysis of One Year of Person Tracking", IEEE Transactions on Human-Machine Systems, Vol. 45, No. 2, pp. 228-237, 2015
- F. Zanlungo, D. Brscic, T. Kanda, "Pedestrian Group Behaviour Analysis under Different Density Conditions", Conference on Pedestrian and Evacuation Dynamics PED 2014, October 2014 (in Transportation Research Procedia, Vol. 2, pp. 149-158, 2014)
- D. Brscic, F. Zanlungo, T. Kanda, "Density and Velocity Patterns during One Year of Pedestrian Tracking", Conference on Pedestrian and Evacuation Dynamics PED 2014, October 2014 (in Transportation Research Procedia, Vol. 2, pp. 77-86, 2014)
- F. Zanlungo, D. Brscic, T. Kanda, "Spatial-size scaling of pedestrian groups under growing density conditions", Physical Review E, Vol.91, No. 6, p. 062810, 2015 [see also the groups dataset]
Inquiries and feedback
For any questions concerning the datasets please contact:
