Deep Neural Networks on a Low Powered Humanoid Robot for Human-Robot Interaction

Alexander Biddulph1, Alexandre Mendes2, Stephan Chalup2, Mark Anthony Wood3

  • 1The University of Newcastle, Australia
  • 2The University of Newcastle
  • 34Tel

Details

10:30 - 13:00 | Tue 22 May | LBP Zone | TuA-LBP.8

Session: Late breaking posters - 1

Abstract

A low powered humanoid robot, named NUgus, was tasked with the job of attracting patrons' attention at an event with difficult indoor lighting conditions. A deep neural network for pedestrian detection was implemented on the robot. As patrons came close to the robot, the robot would wave at them and try to establish eye contact with the closest person.