AI identifies firearms in live CCTV feeds and alerts police
Joe Levy works hard to help avoid another mass school shooting tragedy.
He says his technology, designed to detect a gun using current CCTV cameras, could make an important difference in life-or-death situations in the future.
17 people died in 2018 when a 19-year-old student opened fire at Stoneman Douglas High School in Parkland, Florida, USA.
15 died in the Columbine High School massacre, near Denver, Colorado, in 1999 when a 17-year-old and 18-year-old shot fellow students.
And 22 people died in May of this year when an 18-year-old attacked Robb Elementary School, in Ovaldi, Texas – one of the worst school shootings in US history.
Levi has developed artificial intelligence that uses existing networks of cameras to instantly alert police when it recognizes a firearm, sends them an image, and triggers snapshots and an exact location.
“The first tragedy is the murders of 19 children and two teachers in Ovaldi, Texas,” he told NoCamels. The second tragedy is that the whole event could have been prevented.”
He says his technology would have made a “big difference” if it had been installed on cameras throughout the school.
“He could recognize the shooter with an assault rifle, the moment the CCTV cameras captured it,” he says.
“Unfortunately, because our system was not in place, the shooter was able to enter the building and start shooting.
“Four and a half minutes later, when law enforcement arrived on the scene, they assumed it was just a simple shooting and didn’t take into account the fact that the shooter was armed with an AR-15 assault rifle.
“Because of the lack of situational awareness from the first responders, they were simply overtaken, resulting in the deaths of all of these children and teachers.
“Our system uses existing CCTV infrastructure,” says Levy, who founded 1702aiHeadquartered in Kfar Saba, Central Israel, with offices in Zurich, Switzerland.
And if it sees a handgun, it simply shares a JPEG image, video and alert GPS coordinates directly with law enforcement.
The problem we are addressing is mass shootings in schools in the United States. Take for example the Uvald School shooting, which is one of the worst shootings in recent times, on the Columbine or Parkland School shooting range.
“You have a guy who shoots from outside the school with a machine gun for two minutes. It’s captured by the cameras. He enters the school, it’s been captured by the cameras. He goes to a classroom and shoots for two minutes.
The police arrived and they had no idea what was going on. They hear gunshots, walk down the hallway, and then realize they’ve simply outgrown them.
“It takes 80 minutes and 374 law enforcement officers to neutralize one man. Why? They don’t have what we call situational awareness, and they have no idea what to expect.”
Levy says the technology he has developed is not only super accurate at detecting weapons. It also instantly alerts people on the ground who can take action. And unlike other weapon detection systems on the market, he says it produces only a small number of false positives.
“Even with the latest machine learning and artificial intelligence algorithms, detecting a weapon using CCTV cameras is difficult, because the camera angle and environment are never the same, and the lighting is always different,” he says.
“Some countries can be very cloudy, snowy or rainy. We work exclusively in infrared, which means only seeing the night mode.
“I come from Hollywood, I’ve worked on hundreds of commercials, movies, and reality TV shows, and I even worked for Apple as an image processing expert. So I know what works and what doesn’t.
“In Padua, Italy, we processed 9 million images and got five false positives. Our competitors admitted they had thousands of false positives.
“We have something unique in the way we have learned to detect a weapon. Nobody wants a system where a cell phone will sound an alert, or a banana will sound an alert.”
The 1702ai technology was operated in Oslo, Norway, in Padua, and in a very large transmission hub in Europe to be identified.
We started working with two security companies in America, one of which is insuring about 50 universities. At the moment, we focus exclusively on schools, because our first clients were governments in Europe,” says Levy.
One of the keys to maintaining high fidelity is the constant updates of artificial intelligence. “First of all, AI has to be very accurate, but you have to keep it accurate.
“By sharing data from partner cities, we can prevent the so-called AI drift, which is a very common problem, but one that needs to be addressed.”
AI Drift occurs when technology works almost perfectly in the lab, but not so well in the real world. Levy says it’s essential to prevent micro drift.
“You see AI in a lab in a closed environment and it works really well, but once it’s deployed in another environment, the expectations are not going to be the same,” he says.
A confidence level that might be 96 percent will drop to 80 percent. And then you will have a lot of false alerts. We need to keep our weapon detector very accurate, so we keep updating, like updating an app on your phone. “
Incidentally, the name of the company – 1702ai – was chosen because the first data set it worked with included 1,702 images of real knife attacks.
