Back in early 2020, Grosvenor was working on a solidified product roadmap focussed mainly around the development of our next-generation time clock – the GT8, a range-topping design from the outset, as a feature-rich, highly secure and performant timeclock, building on the success of the original GT10 Android 6 device which had catapulted us into large HCM opportunities, predominately in North America. We were also in the early stages of developing GT Connect, our revolutionary SaaS Cloud-based clock management platform encompassing middleware, PII/Biometric data compliance and data distribution.
At that time, our biometric roadmap was in a steady state, fingerprint technology provided by recognised industry leaders such as HiD’s Lumidigm and Suprema SF range were fully integrated into our timeclock solutions, along with advanced on clock capability for 1:N identification into the tens of thousands. Tried and tested technology, ideal for the HCM industry and adopted as the Defacto solution for biometric identification or verification needs for HCM applications.
The pandemic rolled up in early 2020, and a lot of things changed very rapidly, and our product roadmap was one of the early business casualties of the uncertainty the pandemic wreaked upon us all.
In terms of clock technology, the pandemic impacted one key area more than others, and that was the touchless operation. In a short time, fingerprint technology (the most widely used, trusted and performant biometric technology) was demoted in terms of its functionality to a favoured touchless mode of operation.
Roll on Facial Recognition
Grosvenor Technology had, in fact, evaluated facial recognition prior to this, but conversations with our existing customers led us to believe there was insufficient demand in the HCM market. Fingerprint technology was regarded as mature, reliable and an accepted solution for both identification and verification applications.
And whilst it is fair to say facial recognition had already been around for some time in the market, it is also fair to say it was mainly used in high-end security applications. Entering a secure facility is one thing, but clocking in 100 people starting a shift where every second is either their time or your money is entirely another.
So that was the challenge we faced, and I wish I knew back then what we have learned as a company since that day.
To set the scene, we already had many thousands of devices in the field, so whatever solution we developed, it needed to operate where possible with our existing solutions. Operation had to be as fast and easy to use as the established solutions, and importantly it had to be secure as we were capturing and processing people’s facial data. Finally, and most technically challengingly, it had to be cost-effective; the incremental cost of the technology could not move the overall solution outside of the market price for a timeclock.
As a result of these key design parameters and after some rapid prototyping, we chose to utilise the onboard cameras already present in our GT8 and GT10 clocks and to develop algorithms to recognise a face and apply software-based recognition algorithms for both verify (Card or Pin + Face) or Identify (Face only).
Our initial prototyping also included Mask Detection, and subsequently, we also scoped in thermal screening, both specific requirements driven by the pandemic.
On a bench, we were able to rapidly develop a demonstrable prototype which was performant and user-friendly, able to recognise a face and match 1:10,000 faces in around a second on our GT8 Clock. So far, so good! Then the real work started in terms of translating that to the real world, making it work for our partners in the multitude of environments it needed to.
It became very quickly apparent that the environment and lighting were key factors, and luckily the technology choice we had made of Visible Light and IR Cameras in the GT8 turned out to be the right one. Our GT8 device is equipped with a combined Visible and IR capable camera, augmented with both LED(White) and IR illumination, both of which significantly improve the solution in poor lighting environments.
But technology only takes us so far, and whilst we continue to develop facial recognition software updates and features such as temperature screening, the reality is that the installation and environment really do count. Also key here is the customer’s perception of what face recognition can do – can it recognise the face of someone wearing a large mask, glasses and a hat, for example? The simple answer is no. If there is not enough information (or, put another way, points of minutia available for the algorithm to work effectively), then this is not possible. It may be possible for a verify where it’s a one-to-one match, but generally not suitable for identify.
It’s not just about the technology, as much as it is about effectively communicating the real-world capabilities of facial recognition technology and, importantly, how it can be successfully implemented in differing environmental conditions.
That is why our Professional Services team always lead a facial recognition opportunity to ensure the partner is fully guided through the process. As much as I would like to say this is a point-and-shoot solution, it isn’t. The technology available today at this price point simply cannot do that. However, with good planning and customer engagement, our facial recognition solution is hugely powerful and reliable, processing staff in a touchless mode, quickly with minimal fuss, whilst ensuring end-user information is protected and the end user’s attestation is given, logged and maintained.
Grosvenor, along with our partners, has learned a huge amount about successfully implementing face recognition, and here are a few of the considerations to think about:
What to consider when implementing a face recognition solution?
1. Compliance and attestation to use the Facial data for punching in and out. Number one and, in a legal sense, extremely important. We have developed both on-clock apps and Cloud-based middleware to not only securely collect and process facial data but to allow clear attestation of the biometric owner for the system to collect and process it. We even ensure it is destroyed automatically if the attestation expires or is revoked.
2. Mounting the terminal at a specific height and allowing for the environment around it. As simple as it sounds, if two faces are in the camera view, then which one should it use? After all, it is just a machine!
3. Give people space to use the device without being interrupted by other users (faces!) Putting it in a narrow busy hall is going to be challenging.
4. Lighting is important. The more consistent the lighting is, the better. Try not to point a face recognition device at strong or flashing light sources, and avoid a strong backlight source, as it reduced contrast and, subsequently, the minutia the algorithm uses, reducing accuracy. The best solution is a consistent front face on front lighting.
5. The enrolment process is important, based on the sound computer principle of garbage in, garbage out! Ensure that when staff are enrolled that they are aware that the better the quality of face presentation, then the better their experience will be when using the clock in future. We have a simple guide for exactly that purpose.
6. Allow for large variances in the height of users, you just may need more than one clock or alternative mechanisms for verification.
7. Cleaning of clocks. Smudges and smears on the front glass of a device will reduce the effectiveness of the camera. Sometimes it’s the simple things!
Having said all the above, we continue to actively improve the technology and have a busy schedule of planned updates to face recognition over the coming months. Our philosophy is to continually leverage customer feedback and advancing technology to mistake-proof the solution and improve customer experience, and we continue to invest to make that a reality.
Why adopt facial recognition?
Time is money, and time theft (aka time clock fraud, payroll fraud and ‘buddy punching’) is still prevalent in many workplaces. A recent report suggested that, globally, businesses lose around four-and-a-half hours per employee per week to the practice. In many jobs, that will equate to 10-15% of total working hours – reduced productivity most organisations cannot afford to wear. Other sources put the figure at 2.2% of total payroll costs. The use of facial recognition has evolved significantly and is now extremely reliable and an obvious, simple-sounding solution to tackling buddy punching. To discover our range of HCM solutions, click here.