Computer vision using the human brain

After more than four years of development, the United States Defense Advanced Research Projects Agency (DARPA) has presented a unique computer vision system. Its uniqueness lies in the fact that the computer system uses the EEG signal from the human brain to clarify the recognition.

The Cognitive Technology Threat Warning System (CT2WS) threat warning system consists of 120-megapixel cameras, a radar, a cluster of computers for object recognition and operators with brain scanners.

Any computer vision system produces a lot of false positives. In this case, the 120-megapixel camera produces up to 810 “suspicious objects” per hour. One or several operators are connected to the video surveillance system - and quickly scroll in front of them the suspicious frames that the pattern recognition software has found. At the same time, the electroencephalogram is removed from the operators using head-mounted EEG scanners. If a person sees danger on the screen, it is immediately noticeable on the EEG.
')
The human brain was also surprisingly fast: according to DARPA, they can show the operator up to 10 images per second, without compromising recognition accuracy. When connected to the EEG scanner system, the number of false positives decreases from 810 to 5 per hour.



The photograph shows an electro-optical 120-megapixel camera with a field of view of 120 °. During field trials in the Arizona desert, the system without radar successfully recognized 91% of targets, with radar - 100%.

via ExtremeTech , DARPA