Dataset: People sensing using multiple robots with Velodyne sensors

Overview

Here you can find a dataset of pedestrian measurements taken in a shopping centre using multiple robots equipped with Velodyne HDL-32E LiDARs. The dataset also includes manually labelled characteristics of the detected pedestrians. For more details see the paper listed below.

Dataset description

The following files are included in the dataset:

  • Label data (3 files):
    • gender.csv
    • bag.csv
    • age.csv
  • Point clouds: PointCloud.hdf5

The CSV files contain label data which indicate the bag or gender of the detected people. The files consist of two columns: id and value. The id column contains a unique identifier of each person, whereas the value column contains the label data for gender or bag.

For the gender, the value column has one of two values, M and F, standing for male and female, respectively.
The annotations for the bags are: N = no bag; L = has bag(s) on the left side of the body; R = has bag(s) on the right side of the body; Bo = has bags on both the right and left side of the body; F = has bag(s) in front; B = have bag(s) on the back; U = no label (point clouds are not human’ point clouds, etc…).
For the age, the values C, A and E mean child, adult and elderly, respectively.

PointCloud.hdf5 is a HDF5 file that contains the point clouds for each person at every time stamp. There are multiple groups at the root of the file. Each group contains data for one person, where the name of the group corresponds to the id in the CSV files with label data.

Within the groups there is one subgroup for each instant when the measurement of the corresponding person was taken. The name of each subgroup is the timestamp (Unix time in milliseconds).

The subgroups have several datasets. The name of these datasets indicates the robot id: 0, 1 or 2. (Since people were not observed by all robots at all times, the number of datasets in a subgroup is often less than three.)

The datasets consists of multiple rows, where each row represents one point in the 3D point cloud. The first, second, and third columns contain x, y and z coordinates of the points, respectively.

Download links

PointCloud.hdf5 (629MB)
gender.csv
bag.csv
age.csv

License

The dataset is free to use for research purposes only. In case you use the data in your work please be sure to cite the reference paper below.

Reference

Koki Makita, Drazen Brscic, Takayuki Kanda, “Recognition of Human Characteristics Using Multiple Mobile Robots With 3D LiDARs”, 2021 IEEE/SICE International Symposium on System Integration (SII), pp. 650-655, Iwaki, Japan, January 2021

Contact

drazen (at) i.kyoto-u.ac.jp