What is Facial Recognition?

A computerized image of a face used in facial recognition software.

Facial recognition is a biometric analysis tool in which a computer software confirms someone’s identity using an algorithm and data from an individual’s face. This software uses a series of data points and utilizes a large pool of references in order to:

a.) determine that what it is looking at is a human face

b.) be able to compare this human face to a set of previously registered data points and confirm this person’s identity.

This technology has been a science-fiction trope for decades, and has only recently become a commonplace reality with the advent of advanced artificial intelligence. The same technology Batman may have used to unlock a secret entrance into the BatCave is now a simple procedure we use to unlock our iPhones!

However, as with any new technological advancement, facial recognition has generated its share of controversies. Having a computer scan, register, and store your facial data in a nebulous cloud somewhere just so happens to freak some people out. But hopefully, this article will shed some light on the software and you will be able to form a more educated opinion.

How Does Facial Recognition Work Exactly?

In order to better explain facial recognition, we will discern the difference between facial recognition and facial detection, which are often confused for one another.

Facial Detection

Facial detection and facial recognition are closely related, and are often features found together on more advanced security camera units. In order for a computer to analyze a face, it must first detect a human face. Facial detection is simply the process by which a computer is able to determine whether a human face exists in a particular image or series of images. The computer must distinguish the face from other non-face objects like buildings, cats, dogs, vehicles, etc.

The easiest method is to have the computer first detect a set of human eyes, often the most distinct feature of a human face. From here, the computer will proceed to either consult a set of predetermined rules, reference photos, templates, or a combination of these elements to confirm that indeed, it is looking at a human face. This process is enabled and enhanced using smart algorithms and advanced machine learning, whereby the computer constantly improves its accuracy with the more data it is given.

A group photo of celebrities analyzed by basic facial detection.

Facial detection software will analyze when a human face is present. Often the first step of more advanced facial analysis processes.

Facial detection on its own is very useful for security measures. For example, a security camera can alert you when it detects a person approaching your home, or it can be used for the purpose of counting how many people enter or leave a certain area. They can also be used to activate thermal sensors in hospitals to check your temperature before you enter. Pretty handy eh?

Thermal image with facial detection and temperature sensing features displayed.

Example of facial detection technology paired with thermal sensors to read multiple body temperatures when engaged.

Facial Recognition

After a computer has detected a human face, it then goes through the process of making a unique data map or “faceprint” that marks the particulars of a person’s face. Facial recognition is where the world of biometrics comes into play. Biometrics is the process by which a person is “identified and authenticated based on recognizable and verifiable data, unique and specific” based from your biological features. For example, a computer may mark the distance between your eyes, the delicate structure of your nose, the width of your lips, and several other details that only your mother could love (just kidding, we know you’re beautiful). The algorithm will then store this data and use it to confirm your face the next time the computer detects it.

An illustrated example of a facial data map generated by A.I. software.

An illustrated example of a facial data map generated by A.I. software.

This technology has a wide range of uses, and is being applied to new industries as it grows and develops. Facial recognition can be used to:

  • make airports more secure and efficient by quickly validating passports and confirming passenger’s identities
  • detect fraud on online platforms where a person’s facial ID is used to login to an individual’s account
  • strengthen cyber security as hackers can’t (very easily at least) steal your face and access your sensitive digital information
  • optimize and protect online banking transactions as more and more banks utilize the software to verify your banking activity and ATM usage
  • streamline the patient registration process in healthcare facilities and can detect high temperatures and emotional/physical pain

Accuracy and Controversy

How Accurate is Facial Recognition?

Under ideal conditions, facial recognition software has a 99.98% accuracy. Of course, no technology is 100% without error, and facial recognition is no different. It is the responsibility of those installing such cameras to ensure there is proper lighting, and that the placement of the units are optimal for facial recognition purposes. This doesn’t change the fact that facial recognition accuracy tends to suffer when used in practical, real-world environments.

That’s why confidence measures are an important component to facial recognition software. These measures determine how certain a computer is when resulting in a positive match. Users can set the parameters so that the software will not trigger a confirmed positive match unless its confident measure reaches a certain level of assuredness. These confidence measures are critical when trying to gain the most accurate results from a facial recognition unit.

Why Is Facial Recognition So Controversial?

The ACLU gained headlines in 2018 after releasing a study showing that after using facial technology on members of Congress, 28 members were incorrectly matched with known criminals from a database. The organization used this information to claim that facial recognition software was not ready to be used in a public setting, and could potentially lead to harmful outcomes if put in the hands of law enforcement. Upon further analysis however, it was discovered that the ACLU performed the test with an 80% confidence threshold, far below the recommended setting of 95% for police/legal uses of the software. The manufacturer of the software argued that a repeated test under stricter conditions would have resulted in more accurate output.

The group of Congresspeople the ACLU claimed were falsely matched with criminals by Amazon's Rekognition software.

The group of Congresspeople the ACLU claimed were falsely matched with criminals in their flawed study.

Furthermore, facial recognition has been accused of being inherently biased towards minorities and females. Most of these claims are now outdated however, as the software has improved dramatically over time. Like most programs assisted with artificial intelligence or machine learning, facial recognition works best when given massive amounts of varying data. After all, the software can only operate on the information it is provided. Minorities, by definition, represent a small sample size of the population, and therefore would be difficult for an A.I. to reference. This phenomenon can be seen in Eastern markets, where facial recognition software has an easier time recognizing Asian faces than those developed in Western markets. Simply changing the training set influences the accuracy of the program.

Regardless, this has not stopped many cities in California, Oregon, and Massachussetts from outright banning or restricting the use of facial recognition software from municipal use. Whether it be due to privacy or fear of abuse from authorities, much of the backlash appears to come from a misunderstanding or a misapplication of the software.


As illustrated in this article, facial recognition software is only becoming more accurate, more secure, and more efficient to use. Despite its critics, facial recognition programs seem to have support from most Americans. When attendees of a sports arena were asked how they felt about facial recognition being used to check into stadiums, “80 percent of respondents found it to be a ‘more convenient and engaging way’ to get into the stands.” And this positivity extends even to police and investigative use of the technology. Pew Research Center polled American’s feelings on several different uses of artificial intelligence in society, and found that 46% were supportive of facial recognition technology being used to look for criminals or observe crowds. Twenty-seven percent were against the application, and another 27% were unsure.

With several big tech companies investing in artificial intelligence and more consumers concerned about their security, it’s no wonder why the facial recognition market size is estimated to grow by $763.5 million from 2022 to 2027. From small applications to unlocking our mobile devices to larger ones like capturing terrorists in an airport or finding missing persons in a shopping mall, facial recognition technology has an ever-widening potential to make our world a better and safer place to live. That’s why ENS Security provides top of the line security cameras outfitted with razor-sharp artificial intelligence that can perform facial recognition functions like a dream. Check out what we have to offer by registering with us here.

Akuvox Facial Recognition Intercom and Access Control E18C

The Akuvox Facial Recognition Intercom and Access Control E18C from ENS Security

About the Author: Aaron Avila

Aaron J. Avila is a digital designer, social media marketer, and security professional with ENS Security.

Get Connected with ENS Security

Follow us for more articles, promos, videos, and more!