Alt

SECURITY MAGAZINE presents an interview with DEKI, an Austrian startup that focuses on smart perimeter protection through artificial intelligence and real-time detection. Our questions were answered by the company's product manager, Mr. Tomáš Tichý.

 

Can you briefly introduce your product portfolio to our readers?

DEKI's strategy is to provide clients using camera systems in their business processes with a modern and affordable solution for complete monitoring and inspection automation. A product that is reliable, can adapt to the customer's needs and at the same time allows companies to reduce or maintain their long-term costs.

We have been able to combine all of these requirements to create an AI-powered platform that is simple to use and adds Computer Vision capabilities to conventional cameras.

In our product portfolio we have created neural networks for use in several business sectors, concretely, remote video monitoring for security companies, inspection of usage of protective work equipment within the industry, solutions for the public sector, such as raffic monitoring  in the municipalities and similar.

Specifically, the Deki platform enables automation, from the detection of an object or situation through a common IP camera to the resulting action. In addition to intruder detection, DEKI can also detect various external impulses, such as fire, vehicles and weapons, depending on the user's demand on what type of objects they need to focus on, we can teach our system to detect additional objects in a short time.

 

You deal with security monitoring among other things. How does security monitoring in practice contribute to the recognition of security incidents from false alarms? What is your "product contribution" respecting this?

No matter what solution is used for intrusion detection, if you don't want to pay a person to sit at the premises 24/7, there will always be a trade-off between detection rate and false alarm rate, with the goal to detect only what you want, when you want it, and ignore unimportant events.

Our deep learning based video analysis software by default ignores video noise, sneaking trees, moving clouds, animals, etc. Which can be the cause of false alarms under normal circumstances.

 

In general, what is the success rate in detecting a false alarm?

When we developed a case study for our product with our partner and client Securiton Servis s.r.o, a leader in security monitoring services in Slovakia, our system eliminated 96% of false alarms, saving time and cost of sending security personnel to the property.  

 

Which camera brands does your system work with?

Our software is not limited when it comes to compatibility with camera brands. Any IP video stream can be added to give the cameras the ability for intelligent video analysis, virtual zones and object recognition.

 

Are you using the modern cloud these days, or are your systems so called on-prem?

Since this fall, we can also offer our clients a Cloud solution, but we still see the benefits of using Edge and on-premise solutions for clients who want to have complete data sovereignty and stable real-time detection under any conditions, with or without an internet connection. The handling of sensitive data is already the number one topic and is becoming an important concept in EU legislation that everyone in business wants to be properly addressed.

 

Download PDF