ELI5: facial recognition technology



There are countless things that humans do naturally that take a lot of effort for machines. One such skill is recognising people by their face. Regardless of lighting, accessories or facial hair, we can pick out the people we know by just looking at their face.

With the advancement of artificial intelligence, however, this is one of the skills that computers are starting to replicate. We’re teaching computers not just to see us, but to recognise us.

Now, facial recognition technology is one of the biggest buzzwords in the AI industry. So, here’s the basics of what you need to know about facial recognition.


What is facial recognition technology?

Facial recognition technology is pretty much exactly what the name suggests. It’s technology that recognises faces. But there’s more to it than that. It’s about allowing a computer to recognise people the same way that humans do — by looking at their face.

Facial recognition is a branch of artificial intelligence (AI). It’s part of computer vision and deals with biometrics. In short, facial recognition is the ability of a machine to identify and verify the identity of someone, simply by scanning their face. 

It’s often used for security and surveillance purposes. Facial recognition technology can boost authentication practices, for example. It can help police find missing people. It can also add a new layer of personalisation to our experiences as customers.


How does it work?

There are three core steps for a machine to complete for facial recognition to work: detection, capture, and recognition.

1.      Detection

Before a machine can recognise a face, it needs to notice that there’s a face there to recognise. Face detection is where the technology locates any faces in a digital image or video. It’s a significant challenge in facial recognition technology because human faces are so varied. Plus, they’re often obscured by accessories, make-up, poor lighting and differing angles.

2.      Capture

Once a computer with facial recognition technology has detected a face, it needs to transform it into a usable format. This is where the technology scans the face and translates it into data based on its features. For instance, spacing, size and shape of the eyes, the height of the cheekbones, the nose, lips, ears, chin, etcetera.

3.      Recognition

When the computer has the input face data, it can then complete the recognition. It does this by comparing the input data against another face image or even a database of facial data. So, it can identify the face owner, or verify whether two faces belong to the same person.


What are the concerns?

Facial recognition technology is a divisive topic. On the one hand, it promises a host of benefits. These range from improved security to a valuable helping hand to law enforcement (for instance). But, as with any use of AI technology, facial recognition also invites a host of ethical concerns.

For example, both the training and the use of facial recognition involves the use of personal face data. This means that people may have their faces scanned without consent, or have their facial data stored without their knowledge. Most drastically, some worry that the technology could lead to a surveillance state.

Facial recognition technology also holds some major ramifications in cases where it’s inaccurate. Take law enforcement, where an inaccurate result could cause the conviction of an innocent person. This makes the risk of algorithmic bias a major concern.


Facial recognition technology: judgement pending

Facial recognition, like any other AI function, is a tool. It’s up to us how we use it — and allow it to be used. With the right legislation and regulation, it stands to improve our security, our safety, and our experiences. Used unethically, concerns around the tech are valid.

Of course, this is only a high-level overview. Facial recognition is a broader, (and controversial) subject with many nuances. There’s a host of detailed resources available online if you’d like to dig deeper.

For now, though, you know the face of facial recognition technology — and the main processes and concerns that lurk behind it.


Useful links

What is Moravec’s paradox and what does it mean for modern AI?

Big Brother AI is watching you

Are AI ethics impossible?