Visual perception and its outcoming information plays a crucial role in human being behaviour. As such, robotic systems have been widely exploring such ability for the last sixty years, allowing them to infer on the environment in order to adapt its behaviour. Nevertheless, artificial vision and artificial intelligence were only driven by the growth of computer processing power.

On one hand, vision algorithms applied to the environment acquired content, by vision sensors, enable information extraction. On the other hand, the growing ambition to endow robotic systems with environment awareness requires systems capable of, for instance, detecting and identifying objects. Artificial intelligence, with machine learning algorithms (e.g., neural networks), addresses this issue since it allows the system to learn through acquired data. This cognitive leap aspires to provide the robot a reasonably safe and robust interaction to cooperate with another system (e.g. a human being) in real time.