Kaspersky Lab employees have created a neuromorphic platform that allows a camera attached to it to record more than a million events per second, such as counting grains of sand falling in an hourglass. Andrey Lavrentyev, head of Kaspersky Lab’s technology development department, told socialbites.ca.
“We can analyze more than 1000 frames per second. Only special cameras (neuromorphic ones) can reach these speeds. They do not send an image, but a change in the brightness of each pixel – more than a million events per second. These are otherwise called DVS cameras (Dynamic Vision Sensor) “No computer or network can cope with such a flow of data, because there is no place to store such volumes – they need to be processed immediately. Only the spiking neural network can do this,” Lavrentyev explained.
Experts note that such cameras can be used for autopilots, including drones.
“There is a well-known task for drones to “recognize power lines as obstacles.” It is very difficult for a drone to figure out how to fly around them; drones often hit them. With DVS cameras this task is solved much easier. “In addition, since DVS works in the entire brightness range, thanks to these cameras, drones can also see in the dark: They are not blinded by the sun and can see in the dark,” he said.
Moreover, according to him, such cameras can also be used in production.
“For example, tracking the dynamics of a gas jet in a turbine or processes invisible to the human eye is exactly the task of a neuromorphic platform. This is also important for the industry,” Lavrentyev concluded.
Learn more about what artificial intelligence will look like in the future, what neuromorphic chips are and how they work. report Lavrentyev “socialbites.ca”.
Russian scientists before taught Neural network sleep for optimization.