When considering the installation of facial recognition software in a business or particular infrastructure, the sheer number of providers that are now on the market can sometimes be daunting. Here, facial recognition specialists at Herta look at the key factors and offer advice on how to discern between differing systems, because selecting the correct solution that meets the challenges of the application and that is reliable at the same time, is not always an easy task.
Five important factors to consider before buying facial recognition software:
1) Accuracy
Of course, they all claim to be the most accurate on the market. There are going to be some that even dare to provide matching rates of 99% or 100%. Herta advises to take a “buyer beware” stance on this subject and suggests that such percentages are very difficult to achieve in real environments, since the conditions are different from those of a test environment in which external factors are controllable.
External factors can include lighting conditions, the quality of the images in the database or even the IP camera position. So, in the real world, precision is something that will be achieved by having facial recognition software working in situ. It is very important to be able to get a demo before purchasing any product, and make sure it is a system that has a good reputation among its customers and with existing references.
2) Speed
Although in a video surveillance installation with facial recognition one of the key factors is the level of reliability of the system (that is, a low percentage of false positives), what is also very important is that it is accompanied by an in-system high operation speed. The reason is clear: to be able to detect and act in time. For this, the ideal is that the system can offer identifications in real time.
3) Adaptability
Another indication that the facial identification system is of good quality is that it can be adapted to the existing infrastructure. For example, if cameras are installed along with a VMS to monitor them, the software should be able to adapt and integrate seamlessly. Versatility is key. In addition, it should also be possible to scale the system to other locations and be able to centralise everything in one place.
4) Privacy protection
Most people agree that security is more important than privacy, which is perhaps why many users will be resorting to increasing the security of their business with facial recognition. But it is also very important to protect the privacy of other people, so measures must be taken to ensure that the software does not violate anyone’s right to privacy.
This will depend on the legislation of the country where the application is to be implemented, so the company should be able to offer encrypted databases or image deletion from time to time. Another factor to take into account is the manufacturing origin of the products. It is important to check that the product slated for use does not have a use ban in that country.
5) Final cost
Today many companies minimize the necessary use of hardware, even going so far as to use edge systems to adapt to the client’s budget without affecting the performance of the software. Unlike CPU-based systems, software that uses GPUs to process video streams allows you to analyse more cameras on the same GPU.
Even having ensured that these main points are guaranteed, it is important to always test before purchasing. Optimal demo licenses should last up to 2 months, that should be approximately enough time for a company and its IT managers to “play” with the software and see if it works well with their specific needs and conditions.