HRL Laboratories, LLC announced today that its scientists have successfully mapped HRL’s neuromorphic video object recognition algorithms developed previously under the DARPA Neovision2 program to a spike-domain neural computational framework under the DARPA SyNAPSE program.
The goal of the Neovision2 program was to develop an unattended, stand-alone autonomous object-recognition system that could operate without human supervision under a wide range of day or night environmental conditions.
The Neovision2 algorithms implemented in a spike domain neural computational framework were systematically evaluated using DARPA’s publicly available Tower video dataset and compared with previous implementations for performance and energy usage. This evaluation showed that the spike-domain implementation maintains Neovision2’s high object recognition performance, while being ten times more energy efficient on SyNAPSE hardware.
“As an example, these embedded algorithms will be able to simultaneously process 30 high-definition videos in real-time, while consuming similar power as one 75 Watt incandescent light bulb,” said Dr. Deepak Khosla, HRL Senior Scientist. Future generations of such spike-domain hardware implementations are projected to enable these algorithms to be as much as 500 times more energy-efficient than conventional computers.
“A key feature of these algorithms that makes them accurate and so efficient is that they emulate the visual processing flow of the mammalian brain using spike-based computations,” Khosla noted.
This is an important development enabling low-power video processing onboard mobile and robotic platforms.
HRL Laboratories, LLC, Malibu, California (hrl.com) is a corporate research-and-development laboratory owned by The Boeing Company and General Motors specializing in research into sensors and materials, information and systems sciences, applied electromagnetics, and microelectronics. HRL provides custom research and development and performs additional R&D contract services for its LLC member companies, the U.S. government, and other commercial companies.
Media Inquiries: Michele Durant, (310) 317-5321