Deep CNN and Probabilistic DL Reasoning for Contextual Affordances

Endowing robots with cognitive capabilities for recognising contextual object affordances is a big challenge, which requires sophisticated and novel approaches. In this paper, we propose a hybrid approach to interpret contextualised object affordances from sensor data. The proposed approach combines both Deep CNN networks for object and indoor place recognition with probabilistic DL reasoning for affordance inference. We argue that our hybrid approach can be an interesting alternative in situations where no specific dataset for contextualised affordances exists.