If you read one article about gun detections, read this one. It can help you differentiate a quality provider from from fly-by-the-night providers, whose computer vision models aren’t ready for the critically important task of detecting guns (and knowing what to do automatically if one is detected).
Incorrect Identification of Weapons
AI-based Gun detection using computer vision tends to focus on both object and behavior recognition and can misidentify objects like when a broom handle was detected as a gun in the Lockport school district in New York--the project was cancelled due to facial recognition, privacy, and other issues. When evaluating computer-vision-based gun detection solutions, it’s critical to understand the level of internal expertise, how much time they’ve spent on their algorithms, that the provider has trained their models to distinguish between firearms and laptops, umbrellas, eyeglass cases, among others.
Constant, Unhelpful Alerts
Frequent misidentification of objects as weapons has another effect on the efficacy of any gun detection solution. If staff receives a steady flow of false positive alerts, it’s only natural that they begin to get ignored.
If a system creates numerous false alerts, it’s only a matter of time until those alerts will be ignored. If that happens when a real-life emergency occurs, the results could be devastating.
Privacy Issues with Facial Recognition
When looking into visual gun detection providers, it’s important to note that reputable companies won’t be using facial recognition technology without first and clearly obtaining permission from the people entering buildings. For gun detection purposes, it’s not a necessary feature to include, as the main objective is to detect guns being used by people, not not their faces, specifically. Additionally, if their computer vision models aren’t adequately trained to detect guns, there is also increased potential for discrimination to occur based on how the model was trained for gun detection.
Mass Alerts Expertise Doesn’t Equal Good Gun Detection
A number of “gun detection” providers built their businesses on mass alerts, and have now jumped on the bandwagon to address weapons detections. While it’s true that a good gun-detection system also involves mass alerts, most mass alerts companies don’t have the skillset or acumen to build AI-based algorithms for computer vision, which is a much more complex, harder-to-find, harder-to-develop skillset than building alerting tools.
Instead, it’s logical to work with the companies that already have built years of expertise into their models for gun detection, working with a company that specializes in computer vision and AI products for safety and security that also has two-way communications, mass alerts, automations, and other critical emergency capabilities for robust end-to-end gun detection.
Lack of Integration with Other Key Capabilities
Many gun detection providers simply send mass alerts when a gun has been detected. Unfortunately, that detection doesn’t give organizations enough of an advantage versus an active shooter, who has typically planned out an attack.
In order to minimize harm, it’s critical that the gun detection provider offers deep and flexible functionality to support mass alerts on SMS, Slack, email, and phone and that it will also allow key staff to be immediately informed and give them the tools to make decisions that will initiate downstream activities like calling the police, calling an ambulance, and informing on-site security staff of events as they unfold.
If the system can integrate with building security and access systems to lock down a building, it’s even possible to lock an active shooter in a specific area or room, thereby, preventing harm to those in the building.
Taking it a step further, in the event of an evacuation, being able to determine who is still in the building and who is safely made it outside using surveys or check-in points outside of the building or people counters at entrances and exits also allows local officials to know how many people are still trapped in the building or need help.
In many facilities, the human resources information system, or HRIS, could also be integrated to details which specific people are still missing or in need of help. When an active shooter incident begins, it can take 12 minutes for the police to arrive, so the more information and tools that an organization can layer on to improve outcomes, the better the chances are to reduce harm.
Doesn’t Catch Everything
Computer-vision-based gun detection is a critical tool in the prevention and preparation toolbox, but no single security and safety measure, guards, or technology can catch every weapon. The weapon has to be visible.
Each of the following mass shooting mitigation options has strengths and weaknesses. Consider these options:
- Security guards are humans and the human eye was not created to be “on” and at full attention during entire shifts. This affects both an armed guard at an entrance or a guard watching video feeds on monitors.
- Gunshot detection can be helpful, but it only happens after a shot has been fired and pinpointing locations can be difficult in enclosed spaces, where reverberation can be problematic.
- Physical detection methods like scanners don’t typically pick up plastics. Additionally, backup methods are needed, since they are typically only deployed at key entrances due to cost and staffing reasons, leaving gaps when other secondary doors are opened.
- Gun detection can pick up what the installed cameras can see in real-time and visual detection of guns offers several advantages. They are on 24/7, require no breaks, and are always giving every inch of video 100% attention. This can mean having more time to respond to gun violence or an active shooter situation. They can’t, however, pick up concealed or covered weapons.
No single solution from guards to visual gun detection to audio gunshot detection will ever provide 100% safety in an active shooter event, but when used in combination, each tool provides an additional layer of safety and can potentially give your team time savings and preparation needed to prevent or reduce harm to those in your building.
----