The new FENCE program of the US Department of Defense Advanced Research Projects Agency (DARPA) will help create cameras that can themselves determine what is worthy of attention in the frame.
The Defense Advanced Research Projects Agency (DARPA) has announced the launch of a program to develop a neuromorphic camera, which is designed to improve the efficiency of computer vision by simulating how the human brain processes information.
FENCE (Fast Event-based Neuromorphic Camera and Electronics) development has been entrusted to teams from three companies: Raytheon, BAE Systems and Northrop Grumman. All three are the largest representatives of the US military-industrial complex. Each of the teams will offer its own version of the camera with neuromorphic chips and algorithms.
Modern image processing techniques are becoming more and more sophisticated. The newer ones can capture high-resolution images and track objects with great accuracy, but they do this by processing large amounts of data, which takes time and energy.
The FENCE program must create a camera with a processor that would work according to the same algorithm: all attention should be given to those parts of the image where something worthy of close examination is happening. This will reduce the amount of data transferred for processing and reduce power consumption.
The result can be a fence sensor that consumes less than 1.5 watts of power.