What is Artificial Intelligence based Facial Recognition?

By Leonids Bessonovs
What is Artificial Intelligence based Facial Recognition?

Facial recognition is an integral part of our lives, whether we know it or not. It is used in many situations, from personal security to national security, to marketing and AI learning. Facial recognition uses technology to compose a mathematical faceprint from the features of a human face, which can then go on to do different things. This technology is very young and has a few issues linked to privacy and accuracy, but it is constantly being updated to become better and more capable.


Good morning

The first thing many of us reach for in the morning is our phone; we pick it up and go to unlock it. Throughout the years, methods of unlocking devices have changed - from the simple 6-digit pin, to an elaborate pattern, to your fingerprint, and finally, your face.

Nowadays, many of us have forgotten our original pins. Instead, we now rely on our faces to unlock our devices. But have you ever stopped to think about how these technologies actually recognise you over others?

How would it feel to be broken down and remembered as a set of features saved as mathematical equations? How would you feel if your defining characteristic was the size of your nose or the distance between your eyes? Before getting too self-conscious, let us face the simple truth - this is exactly how our devices recognise us.

“Please present your face to unlock the device.”

Do you recognise me?

So, facial recognition is a way of recognising a human face through technology. The market for such technology is very promising, with its value predicted at $7.7 billion in 2022 (Norton, 2021). This is because it is so versatile - it can be used in device security, criminal surveillance, airport identity verification, marketing using gender, age, ethnicity, and so much more.

Facial recognition works by creating a biometric map from an image of a face. In simple terms, it identifies a face in a picture by searching for eyes, a nose, and a mouth in that order. From this face picture, the AI then measures certain characteristics and geometries, identifying key distinctive features along the way. This includes the distance between your eyes, the distance between your forehead and chin, the shape of your cheekbones, your eye and skin colour, the length of your nose, and so on…

All this information is then passed through various complicated algorithms to be converted and saved as a mathematical formula known as your faceprint - similar to your fingerprint - which is unique to you in most cases.

The faceprint

Depending on the scenario and context of this faceprint, it goes on to do different things.

To unlock your device, the saved faceprint is compared to the faceprint made by the person present in the camera, and if the features are similar enough, access is granted.

In surveillance, things get more complicated, and privacy issues begin to arise. Your faceprint is linked to your identity and is stored in a massive database without your knowledge, along with thousands of other faceprints.

When a security camera first recognises a face in the crowd - a round thing with two dots, a triangle in the middle, and a horizontal line a bit lower - it takes a picture. This picture is then refined to show a proper face with proper features instead of simple shapes, which can then undergo the process of creating a faceprint.

This faceprint is then compared to the faceprints in the database until a few lookalikes are found. This can then allow security to choose the best lookalike, thus identifying the face using the linked identity. In the end, our poor criminal has been recognised on CCTV, identified, and is soon to be arrested…

But what if the identified person was not our criminal?

Problems

Problems can arise when our database grows too big; Big databases need more resources and energy and need to be impenetrable to malicious attacks - the privacy and identity of countless people are at stake. To add to that, with more faceprints to compare to, we want the comparison to be time efficient but, more importantly, accurate. This is a problem because more faces mean more chances of lookalikes, and so this increases the scope for error and could cause the wrong person to be identified. Other issues also point out how easy it is to bypass facial recognition in CCTV by wearing anything on your face - glasses, hat, or facemask. Even more so, our faces do a good job naturally at changing as you age, meaning the faceprint in the database needs to be updated frequently to stay accurate.

Hope

Fret not, as facial recognition is still in its infancy. As of current, the error rate of incorrect identification is 0.08% (Norton, 2021), which is relatively small. This technology is constantly being improved, with algorithms becoming more capable, for better or for worse - you decide. This field comes with many benefits as well as risks and will become an integral part of our lives as we move into the future. Let's use it correctly.

References

Symanovich, S. (2021) “What is facial recognition? How facial recognition works” NortonLifeLock, Available at https://us.norton.com/internetsecurity-iot-how-facial-recognition-software-works.html (Accessed 20th July 2022)

Kaspersky (2021), “What is Facial Recognition – Definition and Explanation”, Available at https://www.kaspersky.com/resource-center/definitions/what-is-facial-recognition (Accessed 20th July 2022)


Author Biography

Leonids Bessonovs is a 17 year old with heavy interests in physics and computer science. With aspirations in Nuclear and Quantum physics, as well as a keen eye for programming and video game development, he spends a lot of time finding creative solutions and answers to various problems and researching topics he finds great interest in, be it botany or quantum computing. As well as that, Leo spends a lot of time outdoors, hiking, sailing, and exploring the world around. Sometimes, it is better to leave your phone behind.

Leonids Bessonovs

Cite this article as:
Leonids Bessonovs, What is Artificial Intelligence based Facial Recognition?, theCircle Composition, Volume 3, (2022). https://theCircleComposition.org/what-is-artificial-intelligence-based-facial-recognition/