The QOPIUS Engine selects the optimal algorithmic content enabling it to address the challenge of “seeing”. We call this intelligence the QOPIUS Retina, which is artificially conditioned on visual processing.

The Qopius Retina draws:

From the human eye's foveation mechanism: the QOPIUS Engine will allocate its resources of computational processing to specific areas of the image that contain the maximum amount of information or novelty. The interactions with the "Prediction lobe" of QOPIUS ENGINE allow QOPIUS Retina to make use of attention-based information processing. 

From the information processing by the human retina: our retina is composed of different types of photoreceptors that are each sensitive to different information in the visual field (color, brightness variations...etc). The QOPIUS Engine reproduces the mechanism by convoluting the visual field with a set of filters.

From the human visual cortex: the visual cortex is composed of different "feature maps" that interact with each other and produce representations of the visual stream. These representations each have a different degree of abstraction.  For example, a low-level map may extract specific spatial frequencies of an image, whereas a high-level map might be more sensitive to the vision of a face or an object). The QOPIUS Retina reproduces this hierarchy of representations of the field of vision.

These three characteristics allow the QOPIUS Retina to deal with visual information robustly and to take into account changes in the environment. 

The highlight of the QOPIUS Retina is that once the structure is conditioned, it learns unassisted and without the help of the engineer. It may also be easily updated when necessary.

Sensorimotor Information are processed in the brain through two distinct channels:

The « What » channel : Semantic Interpretation

The « Where » channel :  Localization, Attention and Vigilance

APPLICATIONS