The growing number of surveillance cameras spread over the territory, as well as the increasing number of mobile devices able to acquire images, video streams or just speed and position, such as smartphones or wearable devices, has lead in the last decades to an increasing interest of the scientific community towards those solutions able to automatically analyse the scene where a person is moving so as to identify events of interest . This interest has been accompanied by a strong pushing of the market, more and more interested in the possibility to run the above algorithms directly on board of cameras, smartphones and wearable devices, or more generally on board of embedded low-cost architectures .

Within this context, the scientific problem, addressed by this workshop, becomes to find out the best possible trade-off between the accuracy required by the particular application at hand and the computational burden needed for its computation . The combination between computer vision algorithms and embedded systems is typically referred to as “Embedded Vision ”.