THOR - people tracks

THÖR is a dataset with human motion trajectory and eye gaze data collected in an indoor environment with accurate ground truth for the position, head orientation, gaze direction, social grouping and goals. THÖR contains sensor data collected by a 3D lidar sensor and involves a mobile robot navigating the space. In comparison to other, our dataset has a larger variety in human motion behaviour, is less noisy, and contains annotations at higher frequencies.

The dataset includes 13 separate recordings in 3 variations:

``One obstacle" - features one obstacle in the environment and no robot

``Moving robot" - features one obstacle in the environment and the moving robot

``Three obstacles" - features three obstacles in the environment and no robot

THOR - people tracks is the part of THÖR data set containing ground truth position of people in the environment, including information about head orientation.  The data are available in three formats:

mat - Matlab binary file

TSV - text file

bag - ROS bag file

MAT files

File - [char] Path to original QTM file

Timestamp - [string] Date and time of the startof the data collection

Start Fram - [char] 1

Frames - [double] Number of frames in the file

FrameRate - [double] Number of frames per second

Events - [struct] 0

Trajectories - [struct] 3D postion of observed reflective markers

Labeled  - [struct] Markers belonging to the tracked agents:

Count - [double] Number of tracked markers

Labels - [cell] List of marker labels

Data - [double] Array of dimension {Count}x4x{Frames}, contains the 3D position of each marker and residue

RigidBodies - [struct] 6D pose of the helmet, corresponds to head poistion and orientation:

Bodies - [double] Number of tracked bodies

Name  - [cell] Bodies Names

Positions - [double] Array of dimension {Bodies}x3x{Frames} contains the position of the centre of the mass of the markers defining the rigid body

Rotations - [double] Array of dimension {Bodies}x9x{Frames} contains rotation matrix describing the orientation of the rigid body

RPYs  - [double] Array of dimension {Bodies}x3x{Frames} contains orientation of the rigid body described as RPY angles

Residual - [double] Array of dimension {Bodies}x1x{Frames} contains residual for each rigid body

TSV files

3D data

File Header

NO_OF_FRAMES  - number of frames in the file  

NO_OF_CAMERAS - number of cameras tracking makers

NO_OF_MARKERS - number of tracked markers

FREQUENCY - tracking frequency [Hz]   

NO_OF_ANALOG - number of analog inputs   

ANALOG_FREQUENCY - frequency of analog input   

DESCRIPTION -  --

TIME_STAMP - the beginning of the data recording

DATA_INCLUDED - the type of data included

MARKER_NAMES - names of tracked makers

Column names

Frame - frame ID

Time - frame timestamp

[marker name] [C] - coordinate of a [marker name] along [C] axis

6D data

File Header

NO_OF_FRAMES  - number of frames in the file  

NO_OF_CAMERAS - number of cameras tracking makers

NO_OF_MARKERS - number of tracked markers

FREQUENCY - tracking frequency [Hz]   

NO_OF_ANALOG - number of analog inputs   

ANALOG_FREQUENCY - frequency of analog input   

DESCRIPTION -  --

TIME_STAMP - the beginning of the data recording

DATA_INCLUDED - the type of data included

BODY_NAMES - names of tracked rigid bodies

Colum Names

Frame - frame ID

Time - frame timestamp

The columns are grouped according to the rigid body. Each group starts with the name of the rigid body and then is followed by the position of the centre of the mas and the orientation expressed as RPY angles and rotation matrix

Reference:

For more details check project website thor.oru.se or check our publications:

@article{thorDataset2019,
title={TH\"OR: Human-Robot Indoor Navigation Experiment
and Accurate Motion Trajectories Dataset},
author={Andrey Rudenko and Tomasz P. Kucner and
Chittaranjan S. Swaminathan and Ravi T. Chadalavada
and Kai O. Arras and Achim J. Lilienthal},
journal={arXiv preprint arXiv:1909.04403},
year={2019}
}