Researchers at Kaspersky Lab have developed a neuromorphic platform paired with a camera capable of recording more than a million events every second. These events are not traditional images but brightness changes at each pixel, enabling the system to tally rapid transitions such as sand grains slipping through an hourglass. Andrey Lavrentyev, who leads the company’s technology development department, described this capability to socialbites.ca.
According to Lavrentyev, the system can process over 1000 frames each second. Only neuromorphic cameras achieve these speeds, he noted. The device does not deliver a conventional picture; instead, it outputs a stream of brightness changes—more than a million events per second. This data flow is so large that standard computers or networks cannot store it; it requires immediate processing. The solution lies in spiking neural networks, which are designed to handle such rapid, event-driven data.
Experts point out that this technology has potential applications in autopilots, including drone systems.
There is a recognized challenge for drones in recognizing power lines as obstacles, which can lead to collisions. With DVS cameras, this problem becomes easier to manage. Lavrentyev noted that these cameras operate across the full brightness range, allowing drones to see even in low light conditions and avoid being blinded by sunlight.
Beyond flight, the technology may find use in industrial settings as well.
For instance, it could monitor the dynamics of a gas jet within a turbine or track processes that are invisible to the human eye. This capability is positioned as valuable for manufacturing and other sectors that rely on real-time observation of fast physical phenomena.
Remarks from the researchers suggest that the future of artificial intelligence will involve more neuromorphic chips and platforms, with ongoing exploration into how such systems can operate at the edge and in real-time for critical tasks. The report on these developments appears in socialbites.ca, providing context on the trajectory of this field.
Historically, other groups have explored neural networks for optimization tasks during sleep states, a line of inquiry that informs current efforts toward efficiency and resilience in intelligent systems. These parallel efforts help frame how neuromorphic approaches might evolve and integrate into practical uses in the coming years.