What Do I Look Like? A Conditional GAN Based Robot Facial Self-Awareness Approach

  • Shangguan Zhegong
  • , Chuang Yu
  • , Wenjie Huang
  • , Zexuan Sun
  • , Adriana Tapus

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

In uncertain social scenarios, the self-awareness of facial expressions helps a person to understand, predict, and control his/her states better. Self-awareness gives animals the ability to distinguish self from others and to self-recognize themselves. For cognitive robots, the ability to be aware of their actions and the effects of actions on self and the environment is crucial for reliable and trustworthy intelligent robots. In particular, we are interested in robot facial expression awareness by using action joint data to achieve self-face perception and recognition, passing a deep learning model. Our methodology proposes the first attempt toward robot facial expression self-awareness. We discuss the crucial role of self-awareness in social robots and propose a CGAN (Conditional Generative Adversarial Network) model to generate robot facial expression images from motors’ angle parameters. By using the CGAN method, the robot learns its facial self-awareness from a series of facial images. In addition, we introduce our robots facial self-awareness dataset. Our methodology can make the robot find the difference between self and others from its current generated image. The results show good performance and demonstrate the ability to achieve real-time robot facial self-awareness.

Original languageEnglish
Title of host publicationSocial Robotics - 14th International Conference, ICSR 2022, Proceedings
EditorsFilippo Cavallo, Laura Fiorini, Alessandra Sorrentino, John-John Cabibihan, Hongsheng He, Xiaorui Liu, Yoshio Matsumoto, Shuzhi Sam Ge
PublisherSpringer Science and Business Media Deutschland GmbH
Pages312-324
Number of pages13
ISBN (Print)9783031246661
DOIs
Publication statusPublished - 1 Jan 2022
Externally publishedYes
Event14th International Conference on Social Robotics, ICSR 2022 - Florence, Italy
Duration: 13 Dec 202216 Dec 2022

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume13817 LNAI
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference14th International Conference on Social Robotics, ICSR 2022
Country/TerritoryItaly
CityFlorence
Period13/12/2216/12/22

Keywords

  • Generative adversarial network
  • Human-robot interaction
  • Self-aware robot

Fingerprint

Dive into the research topics of 'What Do I Look Like? A Conditional GAN Based Robot Facial Self-Awareness Approach'. Together they form a unique fingerprint.

Cite this