Vision systems are becoming ubiquitous in our daily lives. Complex analysis of images from multiple cameras will become the norm in the future, from cars to industrial systems, from smart cities to facility monitoring, aimed at extracting meaningful, context-dependent information.Today’s market is dominated by a combination of relatively simple, fixed function, configurable cameras that stream video to PC-based (and in some cases small embedded) gateways. These systems cannot scale beyond a certain size because of power consumption and the aggregate networking bandwidth required to stream videos to servers, where aggregated video analysis is performed. So the trend for visual analytics functions is that they get executed at the edge of these complex vision systems, e.g. in the cameras themselves.The Cognitive and Perceptive Camera Systems (COPCAMS) proposal leverages recent advances in embedded computing platforms to design, prototype and field-test full large-scale vision systems. It aims at exploiting a new many-core programmable accelerator platform to power a new generation of vision related devices (smart cameras and gateways), able to extract relevant information from captured images and autonomously react to the sensed environment by interoperating at large scale in a distributed manner.Date of approval by ECSEL JU: 11/07/2016.