ZeroEyes uses AI and security cameras to detect guns in public and private spaces

AI has been used a lot for face detection around the world in our surveillance society. But ZeroEyes believes it can detect immediate threats by using AI to detect guns.

ZeroEyes has been rolling out the gun detection video analytics service since 2022. It combines the automated detection of gun-like objects with video analysis by human experts before it sends an emergency message to the place where a shooter might be present — before shootings take place.

ZeroEyes has proven its worth in cases where shooters brandish weapons — sometimes long before they start shooting and police are alerted, said Sam Alaimo, cofounder of ZeroEyes and a former Navy SEAL, in an interview with VentureBeat.

“We are in full blitzscale expansion mode right now,” Alaimo said. “We started at the right time. Our name is known and trusted. We have dozens of clients speaking publicly on our behalf. After operating there for months or years to verifying our technology works the way we say it does. It’s good because every detection like this one means human lives being saved.”

Gun detection in casinos

ZeroEyes has U.S. Department of Homeland Security Safety Act Designation, and it was deployed across nine Muscogee Nation Gaming Enterprises properties, following a successful implementation at River Spirit Casino, the organization’s flagship location, in October 2022.

“Keeping our patrons safe is of utmost importance, and the decrease in response time that ZeroEyes provides is an opportunity to save lives,” said Travis Thompson, director of compliance at the River Spirit Casino in Tulsa, Oklahoma, in a statement. “Those seconds are critical, and now we’re well-prepared and better equipped to enhance safety. Working with ZeroEyes was a unique experience, making real-time adjustments during testing, something I haven’t witnessed with any other company. I’m really proud that Indian Country is taking security and safety so seriously, and our proactive security solutions are benefiting everyone in our communities.”

The River Spirit Casino spans 200,000 square feet and hosts around 10,000 guests daily.

Thompson added, “ZeroEyes has elevated our overall security program to new heights, and we feel safer knowing that it has our backs. The solution operates discreetly, ensuring patrons experience a positive atmosphere at our facilities, while providing unparalleled protection against potential threats in the background without disrupting anyone’s experience. ZeroEyes was the perfect fit for us because of its seamless integration–  we already had the digital cameras, they added the software, we made the connection and from there it’s been a great partnership that strengthened our security measures.”

It has also been adopted by the U.S. Department of Defense, public K-12 school districts, colleges/universities, healthcare facilities, commercial property groups, manufacturing plants, Fortune 500 corporate campuses, shopping malls, big-box retail stores, and more.

Internal research

ZeroEyes pioneered the field of AI-based visual gun detection after finding through research that, in the majority of mass shootings, the shooter reveals their gun well ahead of the incident. In fact, research suggests that 70% to 80% percent of active shooter events involved a weapon that was visible as much as 30 minutes before the first shot was fired.

For example, in the Parkland shooting, the shooter went into the stairwell and sat there for minutes with his gun fully visible, getting mentally prepared. In the 2012 movie theater shooting in Aurora, Colorado, the shooter got suited up in the parking lot beforehand.

ZeroEyes has been compiling proprietary data based on an internal study of casino shootings from 2017 to 2023. Based on 100 shootings during that time, it found that 40% of the time the shooter was able to escape. This usually means that security didn’t immediately know the identity of the shooter and where the incident took place. It also means that responding officers didn’t know who they were looking for and this gave the perpetrator time to evade arrest. 

While a common misconception is that most shootings happen inside businesses and public venues, less than 50% of shootings at casinos happened inside the building. This means that if a gun is detected by CCTV cameras in the parking lot or perimeter of the building, the doors can be locked and business can safely continue inside the casino while security and police work to apprehend the assailant. About 45% of the shootings occur in parking lots.

How it works

ZeroEyes (ZE) is a proactive visual gun detection and situational awareness software platform based on computer vision and advanced machine learning AI. It is layered on existing digital security cameras at schools, businesses, gaming facilities, healthcare facilities and government offices. These are based on modern security cameras, as opposed to older analog cameras where the images were often blurry.

The technology is designed to identify illegally brandished guns and instantly send images to the ZeroEyes Operation Centers (ZOCs), which are staffed by military and law enforcement veterans 24 hours a day for human verification.

Once these experts verify that a gun has been identified, they dispatch alerts and provide situational awareness and actionable intelligence, including visual description, gun type and last known location of the shooter, to local staff and law enforcement as fast as three to five seconds from detection.

This information is invaluable to first responders, who must act quickly with as much information as possible about a potential active shooter. A common issue ZeroEyes sees today when active shooter incidents occur is that first responders lack the situational awareness necessary to locate the shooter, contain the threat, and prevent further loss of life.

As for the tech, the operations center uses off-the-shelf PCs with new centralized processing units and graphics processing units — similar to gaming rigs. It can be on premise or cloud based

In most active shooter events, there are over one hundred 911 calls being made with contradictory reports, creating significant confusion, a “fog of war” and making it impossible to gauge the true nature of the threat. In training scenarios, ZeroEyes has been able to cut response time by nearly two thirds. The company’s goal is to dramatically reduce response time and save lives.

ZeroEyes built its technology stack entirely in-house, in the US. Its proactive AI-based visual gun detection and situational awareness platform was developed using hundreds of thousands of proprietary images and videos, and layers advanced machine learning over existing digital security cameras.

The way to describe its AI detection is that if a human was looking at a security camera and could detect a gun, then its AI would be able to pick up the same gun. Once the AI detects a potential gun, the image is flagged to the operations center, where staff assesses the frame and determines if it is a positive threat.

The staff is primarily military and law enforcement veterans, often those who served in special forces units. They have been specially trained in previous lines of work to understand and identify guns, as well as remain calm and collected during stressful situations. This is the type of expertise and background required to ensure each alert is thoroughly examined within seconds.

The technology is in no way intended to be a replacement for humans, but the reality is that there aren’t enough people in law enforcement or security to cover the 100 million security cameras currently deployed in the U.S. alone.

Origins

Alaimo met cofounders such as CEO Mike Lahiff in the SEAL teams around 2007. They did deployments together and transition out of the military in 2013. Alaimo went to college to get a master’s degree and they met again in the business world. Alaimo was in private equity and he wasn’t satisfied with the sense of purpose that he had in the military.

In 2018, the Parkland shooting happened. Lahiff picked up his daughter one day from school and she had just finished doing an active shooter drill. She was upset by that, and they talked about the security cameras. Hypothetically, they would be useful after incidents, like fistfights, car thefts or mass shootings.

“That’s where he had the idea. How do we take this archaic technology, the security camera? And how do we make it proactive? How do we it so that camera can see a gun before a shot is fired? And in that way, save a life? So that’s the founding story,” Alaimo said.

They found that newer digital security cameras had better resolution and could discern the shape of a gun at a distance.

“We built the algorithm with relatively modern cameras in the last 10 or 15 years in mind, not the older ones,” Alaimo said.

Since they wanted to make sure there were no false positives, they built the recognition software in-house so that it could identify guns while staying away from facial recognition, Alaimo said.

“We don’t want to store biometric data. We don’t even see live screen livestreams,” Alaimo said. “We actually just get an alert when the algorithm says ‘Hey, I think it’s a gun.’”

The company was bootstrapped at the start and it has raised just shy of $30 million to date. Now it has more than 150 people, with a headquarters near Philadelphia.

Detecting guns and toy guns

If it’s a real alert, ZeroEyes can contact security at a school or other place directly via text message and a mobile app. The app will reveal precisely the photo and location where the firearm has been detected.

If it happens to be a toy gun, ZeroEyes can identify that quickly, Alaimo said, and it won’t send an alert about that. But it will notify a place about a fake gun on a lower alert level, such as an email, as that can still be an issue for some places.

“We actually had to do that today with a school in Florida. And they were deeply grateful for it,” Alaimo said.

Getting off the ground

The company took about two years to build its machine-learning algorithm and it took it to the market in 2020, amid the lockdowns during COVID-19. Since no one was going to school, ZeroEyes pivoted to survive and expanded to commercial and government markets.

The first school to adopt it was Rancocas Valley High School in New Jersey. The school let ZeroEyes go there to test the recognition with various kinds of weaponry for about nine months and it eventually became the first client and remains a customer.

To get more experience with recognizing real guns, the company built a “Hollywood grade” lab with green screens outside of its offices in Philadelphia. There, it could test many different kinds of cameras to see if they could capture images and the AI could recognize the guns in different kinds of lighting conditions and surroundings.

“It’s enabled us to basically perfect this algorithm and try to assess any sort of configuration so that we’re never surprised in the real world casinos,” Alaimo said. “We have a strong advocate in the casino space with River Spirit Casino,” where the system works despite colorful surroundings and flashing lights and dark rooms.

He added, “This has been a massive help to us.”

Spreading across the country

As for places with analog cameras, ZeroEyes rarely comes across clients who have them these days, but they are present in older infrastructure or publicly funded institutions or schools in remote locations. In those cases, ZeroEyes asks them to upgrade their cameras first.

Today, ZeroEyes is in 37 states across hundreds of clients and thousands of buildings in the commercial, education and government markets. The company has detected hundreds of guns to date.

“We are, without a doubt, the most widespread gun detection company on Earth,” Alaimo said. “And we have a unique advantage. This is all we do. We have turned out to be one layer in a multilayered security approach.”

The company believes such focus is better than if it did face detection or license plate detection, as the quality level might suffer. If you focus on one thing, you can do it at an A+ level instead of doing multiple things at a B level, he said. In addition, the company can work with many security companies as a result of its single-minded focus on something most of those companies don’t do.

As for rivals, ShotSpotter was in the market earlier with a tech that focused on locating gunshots based on sound triangulation. It rebranded itself as SoundThinking, but Alaimo does not see it as direct competition as it is based on reacting to gunshots after they’ve already been fired, rather than detecting guns beforehand.

Alaimo said the evolution of technology in AI has proceeded rapidly and helped the company develop its service faster. The company is also in the midst of focusing on cloud solutions, rather than on-premises data. One analyst could likely monitor around 10,000 cameras at a time — and perhaps more — based on the alert frequency.

Many of the customers are paying for the service via public funding. The pricing depends on how many cameras will stream data to the operations center. It adds up to perhaps $20 to $50 per camera stream per month.

Saving lives

Alaimo noted one example of a person who brought a gun to a transit platform. Bystanders were looking at their cellphones while someone on a bench pulled out a semi-automatic handgun. The company dispatched an alert and law enforcement arrived quickly and arrested the evidently intoxicated man, who had a fully loaded handgun.

“We just kind of reflected on that one for a second,” Alaimo said. “You cannot quantify the mass shooting, that doesn’t happen. That kind of really brought it home.”

As for the need for this, Alaimo said, “You’re 15 times more likely now to be killed in a mass shooting than dying in a fire in a building. Yet every single building in America has a fire alarm and smoke detectors. So there’s a reason for that. And I think at some point, without a doubt, every camera in the world, whether it’s our technology or not, will have gun detection software. Because we can all agree that if someone has an assault rifle in front of a K-12 school,” it’s an emergency.