mmPrivPose3D: DataSet

Published: 15 August 2024| Version 1 | DOI: 10.17632/pmdr5rgn8c.1
Contributors:
nima roshandel,
, Hoang-Long Cao, Milan Amighi, Hamed Firouzipouyaei, Aleksander Burkiewicz, Sébastien Menet, Felipe Ballen-Moreno, Dylan Warawout Sisavath, Emil Imrith, Antonio Paolillo, Jan Genoe,

Description

Subject: Manufacturing Engineering + Human-Robot Interaction Specific Subject Area: 3D point-cloud data of human movements and gestures captured by a RaDAR sensor synchronized with RGB-D images installed on the base of a robot manipulator. Type of Data: 3D point-cloud, 3D skeleton keypoints; Raw, Labelled Data Collection: 3D point-clouds of humans were collected using an IWR6843AOPEVM RaDAR at a 10 Hz data frequency. 3D keypoints were extracted by the Nuitrack SDK from RGB-D images captured by an Intel RealSense L515 camera at 30 frames per second as the ground truth. Both sensors' frames were synchronized. An offset (3.9 cm in both horizontal and vertical directions) and a 90-degree rotation around the local x-axis were applied to the camera data for spatial calibration and alignment with the RaDAR's coordinate system. Data Source Location: Brubotics, Vrije Universiteit Brussels, Elsene, Brussels, Belgium Related Research Article: Not yet available (Under Submission)

Files

Institutions

Katholieke Universiteit Leuven, Vrije Universiteit Brussel, Imec

Categories

Robotics, Safety, Robot, Radar, Gesture Recognition, Skeleton

Funding

Interuniversitair Micro-Electronica Centrum VZW

HORIZON EUROPE Framework Programme

101070596

Licence