Attentive Stereoscopic Object Recognition

Object recognition in general is still a challenging task today. Problems arise for example from parallel segmentation and localization or the problem to detect objects invariant of position, scale and rotation. Humans can solve all these problems easily and thus neurocomputational and psychological data could be used to develop similar algorithms. In our model, attention reinforces the relevant features of the object allowing to detect it in parallel. Human vision also uses stereoscopic views to extract depth of a scene. Here, we will demonstrate the concept of attention for object recognition for stereo vision in a virtual reality, which could be applied in the future to practical use in robots.