In February 2018 started the first meetings of the DRAAF project, whose aim is to introduce latest Deep Learning advanced techniques in EVITECH's image processing software. Overall project budget is 852 K€, among which more than 740 K€ fundings.

Initiated in 2013 by the image processing research community, Deep Learning techniques have shown in the last years a capacity to overcome performances of former image processing techniques, first for semantic classification (object recognition, Alexnet), but also later for move estimation (DeepFlow), and for several other applications.

If the first prototypes did not allow real time processing, recent progresses show that real time processing can be achieved today, or expected very soon.

Expected benefits are a finer interpretation of observed objects (e. g. car models classification, caddies detection in a shop, or other objects passing through, like items from a purchase order ...), and, in the case of global security applications from Evitech, a recognition of threats in the crowd (such as cars), or in town.

3 partners are funded for this project, two public laboratories, and the EVITECH SME.