Paul Sun, of IronYun, which means iron cloud in Chinese, recently spoke with me about the growing opportunities in visual analytics — thanks to advancements in AI-vision applications, Edge computing and new hardware. I've included highlights from our chat in this brief blog. You can listen to the full discussion here.
Scores of Fortune 500 companies have turned to AI-powered visual analytics to improve security, manufacturing performance and employee safety, among other opportunities. Paul Sun, IronYun president and CEO, says that the growth opportunity for AI vision is massive. Sun cited estimates that as many as 350 million IP cameras exist in the United States alone and maybe as many as 1.5 billion cameras are collecting data globally.
But to realize the opportunity fully, video cameras and the software that powers them must function in far flung and remote locations with varying levels of connectivity and in many different physical location types. That's no easy feat when discussing mega-enterprise operations. Some must process and store colossal data volumes, so it pays to be smart about what is captured, saved, moved or used.
To provide higher video resolutions and fulfill mushrooming bandwidth needs in challenging locales, the capabilities of Edge computing must continue to expand. Sun said, as an example, that for larger companies, just equipping cameras with AI intelligence software might represent a significant hurdle.
"A typical organization, let's say just a midsize company, will typically have anywhere from tens of cameras to hundreds," Sun said. "The bandwidth required to pull that [information] into the cloud sometimes can be very expensive, or it's not even available in remote locations. So Edge computing becomes one of the key solutions in that ecosystem."
IronYun, a B2B AI Vision software platform headquartered in Stamford, Connecticut, offers Vaidio, the company's flagship software platform, to companies mostly focused on traditional access control, safety and security applications. Vaidio features more than two dozen AI video analytics to IP cameras and Sun says the platform is popular for providing accurate results, a wide breadth of functionality and the ease with which it can be deployed.
"The beauty of the software," Sun said, "it's actually a prepaid package, shrink wrapped, plug and play AI video analytics."
The software delivers what his customers need, which are solutions they can "hide" and don't have to think about. According to Sun, his clients don't want to think about the mechanics.
They just want a reliable way to automate the capturing of video data and to do that they need: Edge equipment that provides the necessary processing power, lower latency and ability to operate on a wide variety of platforms. To deliver the required power efficiency, flexibility and performance, IronYun works with AMD CPUs and accelerators.
To help boost efficiency, AMD EPYC CPUs can boost the number of camera streams processed per server and help lower the costs of hardware investments and power needed to get the job done.
When it comes to costs, Sun warned that the costs of processing video analytics are much different from those associated with text-based AI, made famous by ChatGPT. Text requires far less bandwidth.
To illustrate where a good hardware-software collaboration can lead, Sun discussed how AI-analytics software assists humans in monitoring large numbers of security cameras.
"I'm sure you've seen images of security officers sitting in front of 28 screens and trying to identify an intruder," Sun said. "You may be able to do that for 10 cameras, but if you have a thousand cameras, it's not a job for humans, really. So, that industry has been totally disrupted by AI deep learning technology."
Traffic control is another area primed for AI disruption. Sun noted IP cameras and Edge equipment could help large cities synchronize traffic lights and shorten commutes, potentially leading to cleaner air and enormous combined savings in time and money.
For cities to reduce traffic congestion and become smarter, for retailers to cut into the massive losses due to theft and for manufacturers to monitor and improve production lines, video analytics and the supporting hardware must continue to collaborate and innovate.
