Real-time face tracking and recognition on IBM neuromorphic chip

Recent advances in sensor technologies have made it possible to track information from multiple processes and devices on an unprecedented scale. Once data is collected, analytics can be used to generate fact-based insights and guide decision processes. Although sensors can simply collect an environment property and route the collected data to a database or Cloud service, one can also process the data on a gateway device or the sensor board itself. Edge computing, that is, when data gets processed closer to the collection point, can be employed to save network and storage resources, while also making it possible for a device to react to changing conditions on its environment more quickly. On this paper, we present an application for face recognition using TrueNorth, a neuromorphic chip developed by IBM. On this application, the chip has been programmed to identify people in video streams from cameras installed in business meeting rooms and produce a list of room's occupants along with the time windows when the system has detected them in a particular place.