During the Parkland, Florida, school shooting in 2018, the shooter was caught on a security camera pulling his rifle out of a duffle bag in the staircase 15 seconds before discharging the first round. However, the School Resource Officer didn’t enter the building because he wasn’t confident about the situation, and the Coral Springs Police Department had no idea what the shooter even looked like until 7 minutes and 30 seconds after the first round was fired.
If the video system had included technology to recognise the gun threat in real time, alerts could have been sent to the security team. An announcement could have been made right away for all students and faculty in Building 12 to barricade their doors, and law enforcement could have responded a lot faster to a real-time feed of timely and accurate information.
Automatically recognising gun threats
The technology is centred around a CNN that aims to replicate how a human brain would process informationAegis AI offers such a technology, which the company says enables existing security cameras to automatically recognise gun threats and notify security in real-time. The technology is centred around a convolutional neural network (CNN) that aims to replicate how a human brain would process information. This neural network is trained to recognise what hands holding a firearm look like from hundreds of thousands of images in a proprietary data set.
Over time, the system is able to mathematically calculate what a gun threat in a security camera feed looks like with a high degree of accuracy (well over 99% detection accuracy within the first 5 seconds), according to Aegis AI.
“Active shooter situations are often marred by chaos and confusion,” says Sonny Tai, Chief Executive Officer of Aegis AI. “People are in fight-or-flight response and prioritise immediate survival instead of reaching for their phones and calling 911. When the 911 calls are made, callers often provide delayed, conflicting, and inaccurate information, inhibiting law enforcement’s ability to respond.”
Enhances law enforcement response
Tai says Aegis AI helps to clear up that chaos and confusion. “It provides visual intelligence of the location of the shooter, what they look like, what direction they’re heading, and what they’re armed with,” he says. “This real-time information enhances law enforcement response and enables building occupants to make critical decisions that maximise survivability."
|AI methods including deep learning enable high levels of accuracy in detecting weapons in real-time camera footage|
Tai is a Marine Corps veteran and a social entrepreneur who co-founded Aegis AI with the mission of addressing America’s gun violence epidemic. The start of the company stems from Tai’s upbringing in South Africa, where gun violence rates are some of the highest in the world. Growing up, several of his family friends were personally impacted, resulting in a lifelong passion for the issue of gun violence.
In early 2018, Tai interviewed dozens of law enforcement leaders across the country and found that their biggest challenge in gun violence response was the lack of timely and accurate information. Aegis mitigates that challenge and enables both first responders and security staff to respond more rapidly, he says.
More than 99% accuracy in detecting weapons
Aegis is a software-only solution that plugs into security camera hardware and software, including VMSAegis' solution is completely AI-based, says Ben Ziomek, Chief Product Officer. AI methods including deep learning enable high levels of accuracy in detecting weapons in real-time camera footage. “Legacy, non-AI based solutions generally rely on older methods like motion detection, which is not reliable in differentiating between objects such as phones and firearms,” says Ziomek. “Our AI solution lets us achieve more than 99% accuracy in detecting weapons with an exceptionally low false-positive rate.”
Ziomek runs engineering, data science, and operations for Aegis. Before joining the firm, he led teams of AI engineers and data scientists at Microsoft, leveraging AI to identify high-potential startups globally.
Aegis is a software-only solution that plugs into existing security camera hardware and software, including video management systems (VMS). Existing capabilities of a customer’s VMS does initial, basic analysis and then routes the remaining video to Aegis’s processing units for AI analysis. Alerts can then be sent back however a customer wants, including through a VMS. Aegis can also feed information into a PSIM or command-and-control system if requested by a customer.
Equipping customers with AI tools
Our current focus for the company is to get our technology into the hands of as many customers as possible"As an early-stage company, Aegis is pursuing customers through multiple routes, including directly to end-users and via security integrators, distributors, and dealers. They are currently deployed at diverse customer sites including schools, office buildings, industrial facilities, and public buildings, says Ziomek.
“Our current focus for the company is to get our technology into the hands of as many customers as possible,” says Ziomek. “We are working closely with customers across segments and industries to equip them with the tools they need to make their spaces safer. We’re currently working on educating the market on our offerings, as this technology is very new to many security organisations.”
There are no privacy or compliance concerns because Aegis stores no customer data until a weapon is detected, and even then the data is not cross-indexed with any sensitive information, says Ziomek.