Smart devices are all well and good, but wouldn’t you prefer to have real intelligence at your fingertips? As the possibilities of AI continue to develop, we are seeing data storage methods shift away from centralized models and towards distributed intelligence networks, like the intelligent edge. Building on the huge advances cloud computing has delivered, intelligent edge interacts with the cloud but brings the power of AI to the devices themselves. This allows for much more rapid response times, as elements of processing and decision making can be performed on the edge. For facial recognition software, this shift is an essential piece of progress in making it a viable tool.

Facial recognition on the rise

From smartphone logins and personal banking verification through to playful selfie filters, the capabilities of facial recognition have made a speedy transition into the heart of modern UX. Client service functions are key use case in which we’ve seen rapid adoption, and the market is only set to grow – a study published in June 2019, estimates that by 2024, the global facial recognition market will generate $7 billion of revenue. However, for this technology to be rolled out at scale a number of hurdles are still to be overcome.

Security is a universal concern and it’s no secret that human error, fraud and corruption all play a part in any system being able to ensure high levels of reliability, which is why many organizations are exploring other methods of authentication. Biometric based authentication uses AI to recognize unique biological identifiers, such as fingerprints or irises, to create a more foolproof security method. However, within this field, facial-based authentication demonstrated the highest reliability rate when tested against other office entrance applications, boasting a 93% success rate.

As exciting as this sounds, organizations that wish to roll out a facial-based ID system are still facing some hurdles on the road ahead. The quality of the data sets and the APIs which operate the framework can have a major impact on reliability rates. One trial led by the South Wales Police in the United Kingdom staggering 92% error rate when matching football match attendees to a criminal database. In addition, the widespread adoption of the technology at sites such as offices, apartment complexes and public spaces is dependent on some hefty hardware to ensure that software functions in near real-time. As Edge AI capabilities grow more sophisticated, the capabilities of ‘plug-and-play’ IoT devices are unfortunately no longer sufficient. For facial recognition to work in such busy locations, small yet powerful devices with a GPU capable of running AI/ML models, 3D and depth perception cameras, an efficient cooling system and, crucially, connectivity back to the master cloud are required.

Let’s explore how more reliable software can be developed and how the introduction of intelligent edge computing has the power to make this technology a viable solution for security requirements.

Not all facial recognition APIs are created equal

For companies or developers wishing to create their own facial recognition software solutions, there are a host of API products which are ready to go. Microsoft Azure’s Face API package offers services from simple face matching right the way through to emotion recognition, Google’s Vision AI offers a commercially focused suite of packages and Amazon’s Rekognition has a powerful offering on AWS.

However, ready-made APIs do have their limitations, especially when they need to perform on a commercial scale. For example, a company may wish to use facial recognition software to allow customers to access an application securely. Whilst the API might be able to cope with the users face initially – eyes, lips, nose, mouth – it can soon begin to struggle when it has to factor in emotions or different facial angles. Although they have their current limitations, they are under constant development so, in future, we could see more mature and useful out of the box offerings.

In the meantime, laser focused APIs offer a viable alternative for any organization wishing to prioritize reliability in biometric authentication. Laser focused APIs are designed to excel in one use case, for one domain in only one specific solution, but with the absolute highest quality in all conditions. Apple’s Face ID is a great example of laser focused software combined with a powerful Edge AI device in action. The software has one purpose – to authorize screen unlock with only the users facial features. However, it’s been designed to such a high standard that it performs just as well at night as it does in the daytime and isn’t fooled by changes to hairstyle, make-up or clothing.

Satisfying the data needs of facial ID in real-time

One of the fundamentals of new technologies being adopted is their ability to function quickly and robustly. Whilst it’s easy to take it for granted, the fact that users are able to login to apps and services in a matter of seconds is no small feat. Ensuring all of these data interactions can function and validate in near real-time requires some hefty event-driven architecture and data processing capabilities.

Although the cloud is perhaps the first network which springs to mind, intelligent edge computing will also play a crucial role in ensuring facial recognition technology can provide a reliable service. The key difference that intelligent edge brings to the table is the distribution of some of the decision making processes to the device itself. By doing so, latency rates can be significantly reduced and the amount GPU required is also less. Whilst both have their strengths and weaknesses, the most robust solutions harness the power of both to perform local and cloud-based functions for rapid and reliable service times.

Intelligent buildings with intive_EVA

As the potential of facial recognition technology develops, more companies are exploring its implementation to improve both security and usability at office sites. For employees, the enhanced experience of being able to move around their building with automatic verification will no doubt boost employee satisfaction and, what’s more, the company can offer this without any worries due to the granular levels of security control and reduced rate of fraud and human error the software offers.

At intive, our developers have been spearheading some of the latest facial recognition technology with a specialized API system and the staff working at our Wrocław office are benefitting from the fruits of this labour with the launch of intive_EVA in October 2019.

Those days of forgetting your ID badge and being held up at reception to explain the situation are long gone – our software allows registered staff members to move in and out of the intive offices with ease, with no loss of security for the company. intive_EVA uses image processing algorithms that interact with a 3D camera to guard security of the entire system in Wrocław. During tests, we were able to demonstrate that EVA has a 100% efficiency rate – not once did the device allow an unauthorized person access to the building. We achieved this reliable quality because we embraced the power of intelligent edge.

Achieving 360 Security Solutions

As highlighted, some of the main challenges facing the future biometric identification systems have focused on the ability to reliably demonstrate better than human verification rates and feasibly scale computational models and hardware needs. As we begin to overcome these and prove what 100% reliability looks like, a new generation of possibilities become available for the Smart Buildings of the future.

Similar Insights:

intive and NETZSCH Group Come Together For the Future of Smart Manufacturing
5 Aug 2020
Next-gen shoppers demand next-gen payments
9 Jun 2020
Krzysztof Machelski on AI & ML in Financial Services
28 May 2020

How can we help you?
Let’s talk.

Get in touch

Do you see yourself
among us? Great!

Join us