From visual patterns to semantic description: A cognitive approach using artificial curiosity as the foundation

In this article, we present a cognitive system based on artificial curiosity for high-level knowledge acquisition from visual patterns. The curiosity (perceptual curiosity and epistemic curiosity) is realized through combining perceptual saliency detection and Machine-Learning based approaches. The learning is accomplished by autonomous observation of visual patterns and by interaction with an expert (a human tutor) detaining semantic knowledge about the detected visual patterns. Experimental results validating the deployment of the investigated system have been obtained on the basis of a humanoid robot acquiring visually knowledge from its surrounding environment interacting with a human tutor. We show that our cognitive system allows the humanoid robot to discover the surrounding world in which it evolves, to learn new knowledge about it and describe it using human-like (natural) utterances.