In AI, the edge is at the center of everything right now.
Edge AI processing is moving center stage as more devices are starting to incorporate the high-performance, extremely energy-efficient chips needed to process massive amounts of data coming in from sensors. Security cameras are leading this charge, in some cases integrating not only state-of-the-art vision sensors, but also becoming capable of accepting data from radar sensors and others, making them an early proving ground for sensor fusion.
The market for AI-accelerated cameras is expected to increase at a compound annual growth rate of 23%, according to a Yole Intelligence report from last year that described security imaging as “changing drastically.”
With all of that in mind, it should come as no surprise that ISC West, the Spring event for the International Security Conference, has become an important stop for the AI processor companies, such as Ambarella, Hailo Technologies, and others, all of them looking to get their new processors designed into the latest security cameras.
As ISC West kicks off this week, Ambarella, will be showcasing its recently announced CV72S 4K, 5nm AI vision system-on-chip.
That processor leverages the company’s CVflow 3.0 AI architecture to support the fusion of camera vision and radar data from multiple sensors, and in combination with 5-nm process technology low power consumption of less than 3 watts. The CV72S promises 6x greater performance than Ambarella’s previous processor, and according to Jerome Gigot, senior director of marketing at Ambarella, runs the latest transformer networks, which outperform convolutional neural networks for most vision tasks.
“Our customers in the industrial space want more AI, they want to see more and see longer distances, and what's happened the last couple of months now is that they all want to support transformer [neural] networks,” Jerome Gigot, senior director of marketing at Ambarella, told Fierce Electronics. “Everybody knows transformer networks now because of ChatGPT, and everyone wants that capability.”
The CV72S also can run Ambarella’s Image Signal Processing technology for neural network-enhanced 4K, long-range color night vision and HDR at very low lux levels with minimal noise and no external illumination, while leaving headroom for concurrent neural network processes like person tracking and mask detection.
“Ambarella’s CV72S fits perfectly into the ongoing trend for security cameras to integrate more and more AI capabilities at the edge via camera processors, while maintaining a low power and thermal budget,” according to Florian Domengie, Senior Technology & Market Analyst, Imaging at Yole Intelligence, who was quoted in an Ambarella press release.
Gigot said cameras with the new SoC are likely to be used in smart city applications and similar scenarios, and that the ability to leverage sensor fusion with radar data will come in handy at night and during times when weather might impair vision sensors. He added that radar can add data on speed, distance, direction of movement and other factors that can be particularly helpful in smart traffic applications.
Gigot said, “You cannot rely on just visual data because what happens if it's raining, right? Radar is a great complement to visual because you can use different modalities to see what’s happening.”
For its sensing capabilities, Ambarella’s processor relies on the company’s Oculii virtual aperture imaging AI radar technology, which the company claimed features 10-100x better radar resolution than other radar solutions. Ambarella acquired Oculii in 2021.
RELATED: As foundry market falters, Samsung boosted by Ambarella