PhD defence by Xiaodong Duan
23.01.2019 kl. 13.00 - 16.00
Machine Learning Methods for Selected Vision Tasks in Social Robots
Social robots have recently gained tremendous attention due to a broad range of applications in various environments, e.g., nursing homes, schools, private homes and shopping malls. All of these applications require the robot being able to establish a natural interaction with its users. In human-human interactions, a large portion of them are conducted through vision. Humans are able to recognize people they know and even the kinship between different people, all based on their faces. Furthermore, humans are good at proposing questions through observing surroundings in order to initiate a conversation with others. If equipped with these abilities, social robots will be more capable of having a natural interaction with its users and be more proactive, than without.
In this talk, Xiaodong Duan will present several machine learning methods for selected computer vision tasks in Human Robot Interaction (HRI), in particular, face recognition, kinship verification and image-based question generation. Firstly, feature learning methods for face recognition and kinship verification will be discussed. Secondly, an answers-guided deep learning method for image-based question generation will be presented. Finally, building of two interactive multi-modal social robots will be introduced.
Associate Professor Lars Bo Larsen, Aalborg University, Denmark (Chairman)
Professor Robert Jenssen, University of Tromso
Professor Lars Kai Hansen, Technical University of Denmark
Professor Zheng-Hua Tan, Aalborg University, Denmark
Associate Professor Ove Andersen, Aalborg University, Denmark
After the defence there will be a reception at Fredrik Bajers Vej 7, B4-211
Free of Charge
Signal and Information Processing, Department of Electronic Systems
Aalborg University, Fredrik Bajers Vej 7A4-108