Eagle Eye Networks acquires AI leader Uncanny Vision

AUSTIN, Texas, and BANGALORE, India—Global cloud video surveillance leader Eagle Eye Networks announced the acquisition of artificial intelligence (AI) innovator Uncanny Vision, including the company’s research and development capabilities and a new regional office in Bangalore, India.

Eagle Eye Networks CEO Dean Drako told Security Systems News that Uncanny stood out among other AI companies because they are “entirely and completely focused on AI; It’s not an afterthought,” he said. “I believe they may be one of the first companies applying neural networks and machine learning to video surveillance. And I have always maintained that AI for video surveillance is drastically different than AI for other types of video.”

Another differentiator is the Uncanny team’s knowledge level. “Most of the AI teams that I have encountered have been smart people … but if you tried to talk machine code or assembly code, it was a foreign concept – they really had no idea how the libraries of neural networks are actually implemented to hardware. The team at Uncanny are very different – they really understand how the code can be implemented and optimized to run very fast on the GPUs and TPUs that are available today.”

As only the second acquisition in Eagle Eye’s history, “This is a key one,” Drako noted. “Very important for us and our strategy, especially when you think about the compute capability in the cloud, which is far easier for our customers to deploy (just click a button) and at far lower cost because of the economies of scale of doing it in the cloud.

“Eagle Eye will be the first company in the world to be able to provide that level of functionality, and actually the only one that can do that because we are the only ones who send all of the video to the cloud.”

Future of AI

Drako is very excited about the current and future application of AI within security.

“I think we are heading into this whole realm of AI greatly impacting video surveillance,” he explained, giving a great example of how Uncanny’s AI is making a difference within license plate recognition (LPR), outperforming dozens of other LPR companies during testing.

“If you put these other companies in uncontrolled circumstances where they are in a parking lot and not focused on a spot where cars are coming in – just in that general direction – most of those we tested dropped down to 80-90 percent accuracy,” Drako noted. “But Uncanny just knocked it out the park and were always 98- to 99-percent accurate, no matter what we threw at it.”

Drako gave an example of a van with graphics around the license plate that was throwing most LPR companies off. “The Uncanny code didn’t even flinch and totally ignored the graphics, zooming in on the plate,” he explained. “It didn’t get confused, whereas it threw off almost every other one we tested.”

Another area of AI that Drako is excited for is search functionality.

“It really boils down to object categorization, so when you have a lot of video coming in, you want to categorize things – that is a person, a car, etc. – and you want to do this on all of the video, so if you need to return later to the footage, you can say, ‘show me all the red suitcases on Tuesday’ or ‘anyone with a baseball cap’ during a certain time period on any of the cameras. For example, out of the 14,000 cameras, show me all of the ones that match that criteria – a “guy with baseball cap” and “a red suitcase” – so you can be far more productive in your forensic work.”

Drako noted that this “efficiency of implementation” is what makes Uncanny stand out among a sea of AI companies. “If you want to do that level of categorization on every video all the time, you need to have a ton of servers with GPUs running on site to deliver that level of AI to whatever number of cameras you have. That isn’t practical for 99 percent of customers, so the key is being able to provide that level of search functionality in the cloud without a significant cost ramp up. The only way you can do that is if you can implement these algorithms efficiently on the hardware, and the Uncanny team knows how to do that.”

All 60 Uncanny Vision employees will be retained, and Eagle Eye plans to expand the Bangalore office. In addition, Uncanny Vision co-founders, Ranjith Parakkal and Navaneethan Sundaramoorthy, have joined the Eagle Eye leadership team. Eagle Eye is committed to supporting Uncanny Vision’s current customers and will continue to build its global infrastructure to provide the very best 24/7 support to its valued customers around the world.

“We share the Eagle Eye team’s vision to deliver advanced, cyber secure AI cloud video surveillance offerings that transform video surveillance for businesses around the globe,” Parakkal said.

The deal also accelerates Eagle Eye’s plan, announced in November 2020 (when Eagle Eye raised funds from venture capital firm Accel) to dramatically reshape video surveillance. Uncanny Vision’s deep learning algorithms enable recognition, identification and prediction, improving business operations, customer service and site safety. Some applications include smart parking, retail analytics, gate security, toll automation, smart cities, ATM monitoring, worker safety and perimeter security.

Overall, Drako is bullish on the rest of the year, noting that Eagle Eye should see continued growth, right around the 85 percent growth rate the company saw in 2020.