Integrating Context-Free and Context-Dependent Attentional Mechanisms for Gestural Object Reference

An augmented reality human–computer interface for object localization in a cognitive vision system

“…This requires an accurate selection of the object and segmentation from the background to be adequate for the fast object learning component [14]. Although this can be achieved by only defining the correct 2D boundaries of the object view, the correct 3D position of the object is required when additional stationary cameras with pan tilt units focus the referenced object from different viewpoints or when several mobile users want to interact.…”

The visual active memory perspective on integrated recognition systems

“…On the one hand, a VPL classifier as introduced by Heidemann et al [24,25] is applied. This approach was motivated by biological information processing principles which are believed to underlie early visual processing in the brain.…”

Multimodal interaction in an augmented reality scenario

“…It is based on a neural architecture called "VPL" described in detail in earlier work, see [12] and references therein. In short, the VPL-classifier consists of three processing stages: The first two layers perform feature extraction by means of local principal component analysis (PCA), which is implemented in the separate steps of clustering the raw input data by vector quantization ("Vlayer") and a subsequent local PCA ("P-layer").…”

“…a cone-shaped activation in the direction of pointing, which highlights the nearest maximum of M . A detailed description of the underlying saliency algorithms, the integration to a single map M , and the selection of pointing targets can be found in [12].…”