A system for Gender Recognition on Mobile Robots

The combination of artificial intelligence and robotics opens the way to disruptive future developments in the industrial and collaborative robotics. The recent advances of the deep learning technologies materialize the possibility to provide a robot perceptive and reasoning skills and, consequently, the capability to autonomously interact with a human. In this paper we ride the wave of intelligent robotics by designing an autonomous robot able to recognize the gender of the customers in a shopping center and to interact with them proposing customized advertising and promotional material. We train two well-known Convolutional Neural Network architectures to recognize gender from face images. In order to run them in real time we extend the computational capabilities of a social robotics platform with an embedded parallel computation accelerator. The experimental analysis, carried out on video sequences acquired in real scenarios, demonstrate the suitability of the proposed platform for the considered social robotics application in terms of both latency and accuracy.