Visual Learning of Group Interaction Dynamics - Oya Celiktutan

Описание: Visual Learning of Group Interaction Dynamics
Abstract: Social interaction is a natural part of human life. We interact with others in many ways, ranging from greeting a colleague while passing the corridor to participating in business meetings, and mingling at a social event. Therefore, there has been a significant effort in recent years to automatically recognise and analyse such interactions from videos, which opens up a myriad of applications, including understanding multimedia, developing analytics tools and embodying interactive behaviours via robots. However, it remains a challenging problem and there is still a large room for exploration. In this talk, I will present our ongoing work on the visual analysis of group interaction, involving groups of two (dyads), three (triads) or larger groups. In particular, I will show how the analysis of group interaction can be useful for developing diagnostic tools or learning analytics tools and making robots to navigate and interact with people in crowded social spaces.


Speaker Bio: Oya Celiktutan is a Lecturer in the Centre for Robotics Research, Department of Engineering, King’s College London, where she is the Head of Social AI & Robotics Laboratory. Her research interests centre on computer vision and machine learning applied to human behaviour understanding, social signal processing and human-machine interaction. Particularly, her team focuses on developing smart algorithms to learn multimodal representations of human behaviour from data purely and integrate such models into the perception, learning and control of real systems such as robots. Her team’s work has been supported by UKRI Engineering and Physical Sciences Research Council and Toyota Motor Europe so far.