Hailo Technologies, a Tel Aviv, Israel, company that is one of many pursuing increasing opportunities in the market for edge AI processors, this week announced the new Hailo-15 family of high-performance vision processors, designed for integration directly into intelligent cameras used for smart cities, security, retail and other applications, and aiming to deliver more advancing video processing and analytics at the edge.
While some companies targeting the edge AI market are focused on AI accelerators and a handful of others developing system-on-a-chip (SoC) architectures to be incorporated into new edge devices, Hailo is doing both. The company already offers the Hailo-8 AI acceleration processor, a neural processing unit, and now the Hailo-15 SoC vision processing unit (VPU) is bringing deep learning video processing and greater AI performance into cameras without blowing out cost and power consumption, according to Hailo CEO Orr Danon.
“Instead of being an accelerator the Hailo 15 is a single chip and SoC solution for smart cameras,” he told Fierce Electronics. “It's basically some of the same technology that we introduced in Halo 8 though, structured with different parts of the chip performing different parts of the AI processing task, and incorporating this now into a more complex system that takes the end to end vision pipeline from sensors to connectivity, with a strong emphasis on performance.”
He added that an SoC is more appropriate than an AI accelerator if a camera developer wants to deploy AI processing directly into the camera. “I think we're bringing the level of performance that people are used to seeing from the Nvidia Jetson lineup into something that could live in a camera,” he said.
The Hailo-15 VPU family includes three variants — the Hailo-15H, Hailo-15M, and Hailo-15L — to meet the varying processing needs and price points of smart camera makers and AI application providers. Ranging from 7 Tera Operation per Second (TOPS) to 20 TOPS, the latter able to support multiple sensors and processing tasks in parallel, this VPU family enables over 5x higher performance than currently available solutions in the market, at a comparable price point, Orr claimed. All Hailo-15 VPUs support multiple input streams at 4K resolution and combine a powerful CPU and DSP subsystems with Hailo’s field-proven AI core.
In a nod to what could be a rapidly emerging trend in this market, Hailo also designed support for transformer network architectures in the Hailo 15.
Hailo will be showing off its Hailo-15 family March 28-31 at ISC West in Las Vegas, a showcase for new security camera technology. It certainly will not be the only edge AI processor company bringing new products and capabilities to the event, as Fierce Electronics has learned of at least two other similar announcements that will go public either at the event or in the days prior to it.
That should come as no surprise to close observers of the market, as a variety of processor companies are chasing emerging edge AI opportunities in fields such as security cameras, automotive, smart cities, smart factories, retail, and more. Firms developing these processors or technology to run on them include Hailo, Ambarella, Brainchip, Perceive, Prophesee, Quadric, Deci, and others, some of them with slightly more famous names – Nvidia, AMD, Intel, NXP, etc.. (If you want to get to know almost each and every company playing in this sector, check out the Edge AI + Vision Alliance members list.)
All that activity is eventually going to lead to some market consolidation, Danon observed, with bigger players perhaps building their own technology stacks by absorbing the smaller ones. It remains to be seen what role Hailo could play in that consolidation. The company announced a $136 million Series C funding round back in October 2021, and had banked more than $224 million up to that point.
“We are well funded at this point,” Danon said. “We will need to raise money eventually, but lucky for me, not right now."