Psychologically motivated scene description framework
Humans are able to effortlessly describe scenes at great length by composing together a small list of concepts which can be learned from a few examples. However, this task remains challenging for machine learning models. This work proposes a framework for scene description that addresses these issues. This model is able to learn concepts and apply them to unseen contexts in real time, from no prior information. Inspired by work in cognitive science, this model aims to facilitate understanding between human and machine agents in terms of visual scenes.
Keywords
Computer Vision, Artificial Intelligence, Cognitive Science
Staff
[Kasim Terzic]{kt54} and Daniel Koudouna