Facial recognition is everywhere. What once started as an attribute specific to sci-fi movies is now a part of everyday life: we rely on facial recognition every time we unlock our phones, tag friends in a Facebook post, or go through customs.
The numbers are only proving the ubiquitous nature of facial recognition. In 2020, the global market of facial recognition software was estimated at $3.72 billion, and it is expected to hit the mark of $11.62 billion by 2026, registering a CAGR of approximately 21%.
Along with gaining market volumes, facial recognition algorithms are becoming more sophisticated. Since the outbreak of COVID-19, facial recognition software vendors have upgraded their algorithms with new features, such as recognizing faces with a mask on.
At the same time, however, facial recognition technology has been subject to numerous ethical concerns amplified by reported uses of the technology for racial profiling and protester identification.
With more organizations turning to AI technology solutions for businesses, questions arise: how does AI-based facial recognition work, what are the benefits of facial recognition, and how does one implement the technology ethically? We’ve explored these points in this blog post.
The meaning of the term "facial recognition" is quite intuitive. The technology uses biometric algorithms to map, analyze, and confirm the identity of a face on a photo or a video. Although every facial recognition solution (which often rely on proprietary algorithms) operates differently, we can distill the facial recognition process down to the following three steps:
Detection refers to the process of locating a face in an input image. So, each face is placed into a bounding box. To complete this stage, the facial recognition algorithms are first trained to learn what a face looks like on various data entries. Analysis refers to mapping out the features for each face. This is done by measuring the distance between the eyes, the nose, and the mouse, as well as identifying the shape of the chin. Those distances are then combined and converted into a unique set of numbers — the so-called faceprint. Recognition refers to actually determining a person’s identity in the input photo. In some applications, this stage is replaced with categorization. In such cases, the algorithms do not confirm a person’s identity but label the person as belonging to one of the distinct groups, for example, by gender or age.
When incorporated in a hospital’s video surveillance system, facial recognition can simplify patient check-in, freeing hospital workers and patients from paperwork and preventing human error. By “looking” at a patient, a facial recognition system can verify their identity and insurance data, thus speeding up the admission process, laying a basis for a personalized experience, and preventing fraud.
Biometric technologies, including facial recognition, can also be used to verify the identities of surgical patients, identify patients who are unaccompanied by a medical worker, and track people entering and leaving the premises to prevent security threats.
The applications of facial recognition for patient tracking have the highest patient acceptance rates. A research published on Plos One states that almost 66% of patients find it acceptable for hospital systems to scan their faces for identity verification.
Diagnosing genetic disorders
Facial recognition can help diagnose rare genetic disorders, especially those with mild symptoms. One example of such assistance comes from Delaware’s Nemours Children Hospital, where clinicians struggled to diagnose a patient experiencing unusual symptoms. The clinicians turned to Face2Gene, a face recognition-powered mobile app one of the hospital’s employees, Dr. Karen Gripp, helped develop. The application scans a patient’s photo, mapping their face with 130 landmarks, and uses machine learning to match the detected facial characteristics to those of rare genetic conditions. As a result, the application generates a list of potential diagnoses, each with a probability score. With the assistance of Face2Gene, the medical staff identified a rare case of Wiedemann–Steiner syndrome.
Face2Gene is not the first app that relies on facial recognition to diagnose rare diseases; it is though one of the most popular ones. The developers of the app report that the product has been used to evaluate 250,000 patients and helped identify 7,000 conditions.
Preventing the spread of COVID-19
With the pandemic outbreak, facial recognition systems adopted a wider use — they are now being used to track COVID-positive people who must stay home. A mobile app equipped with facial recognition features asks a quarantining person to take a selfie, then confirms their identity to make sure they abide by the self-isolation rules.
And in South Korea, the government is planning to take facial recognition-enabled tracking further. Despite rising privacy concerns, the government is about to launch a pilot wherein facial recognition software would analyze footage from more than 10,000 CCTV cameras in Bucheon, one of the country’s most densely populated cities. The goal is to track infected people, pinpoint who they communicate with, and state whether they take precautionary measures to prevent the virus from spreading. Similar systems have already been rolled out in China, Russia, India, Poland, Japan, and several US states.
Facilitating mental therapy
Facial recognition helps track patients’ mental health patterns and behaviors. For example, the software can interpret the emotional state and improve the safety of patients prone to risky behaviors, such as removing a breathing tube.
The technology can help doctors cope with stress, too. According to a Medscape report, 44% of physicians feel burned out, 11% are colloquially depressed, and 4% suffer from clinical depression. Facial analysis is reported to identify these conditions and nudge medical workers to take stress-relieving measures.
People with special needs, too, can benefit from facial recognition. For example, researchers at Stanford University developed a facial recognition system that runs on Goggle Glasses. It analyzes people’s facial expressions and prompts the wearer with respective cues, like ‘anxious’ or ‘happy.’ The developers say their solution can help children with autism recognize facial expressions and improve the quality of their social interactions. The trials showed that children who relied on facial recognition software along with standard care showed improvement in socialization as opposed to the control group, who only received traditional care.
Checkout-free software solutions
When brick-and-mortar stores reopened following the lockdowns, customers, wary of excessive interactions and touch, were still reluctant to shop offline. To lure consumers back in, retailers have turned to recognition technologies and specifically facial recognition-enabled contactless payments.
Contactless payments technology relies on machine learning algorithms to let customers pay for the goods by simply scanning their faces. Since goods are RFID-tagged and a customer’s faceprint corresponds to a specific payment method in a database, it is no longer necessary to manually scan the purchased items or communicate with a cashier.
Face recognition-based checkout has long been a thing in China. Even before the pandemic, more than 60% of purchases made on a Singles’ Day (a Chinese equivalent of Black Friday) were paid for by taking a selfie.
In the US, on the other hand, the technology is still taking off, with the first nation’s facial recognition payment system being rolled out in Pasadena as recently as 2020.
Integrated into a store’s CCTV cameras, facial recognition technology can help retailers reward loyal customers without interrupting their buying experience. The moment a loyalty club member enters a store the facial recognition system identifies the customer and, say, rewards them with a personalized discount or informs them about deals or products they might be interested in.
Cali Group was one of the first US companies to have rolled out a facial recognition loyalty program. They equipped their restaurants with AI-powered self-service kiosks that identify registered customers and activate their loyalty accounts as soon as they approach the kiosk. The software powering the kiosk may prompt customers to order their favorite meals. Payments are handled via facial recognition as well.
Personalized shopping experience
A similar approach is taken to create personalized customer journeys. For example, knowing how much time a particular customer spends in a store helps tailor future experiences to their preferences. And analyzing the history of one’s purchases, retailers may nudge buyers with push notifications advertising products similar to those they have recently purchased. Tapping in other data, such as how often a customer pops in a store, when they last made a purchase, or what they buy more frequently, can help consultants adjust their services to a particular customer and provide more personalized assistance.
Store security and fraud prevention
According to a recent research, almost half of the surveyed buyers would like to see facial recognition implemented in stores if its goal is preventing cases of shoplifting.
Systems that prevent shoplifting are usually targeted at identifying repeat offenders whose photos are already stored in a database. A facial recognition system, thus, does not attribute any personally identifiable information to the face of a shop visitor but searches for its equivalent in the database of known offenders.
Strengthening campus security is one of the essential applications of facial recognition in education. The technology has gained extra importance as the incidents of shootings have prompted school administrations to take more advanced precaution measures to prevent tragedies.
Facial recognition systems, thus, were deployed across thousands of schools across the country. Here’s how they may operate in a school setting. A face recognition system analyzes the faces of people entering or navigating the campus and compares them to a database of authorized individuals, including students, the school’s current staff, and parents, to establish their identity. Suppose a person is not found in a database or matches the identity of an unwanted person, for example, an expelled student or a former employee. In that case, the system immediately alerts the security and automatically denies the visitor entry into the campus area.
Along with spotting unauthorized visitors and tracking flagged individuals around the site, modern facial recognition solutions come up with additional features, like object detection (used for identifying gun-shaped objects as a rule).
Tracking attendance used to be a lengthy and tedious process that, despite a fair amount of time spent at the beginning of every class, leads to inevitable gaps and omissions when conducted manually. To fix that, educators have started turning to AI-powered educational solutions. This way, facial recognition applications offer a faster and non-disruptive way of tracking who is present. Not only does it save precious learning time, but it also allows curriculum designers to create more accommodating learning environments and perfect class scheduling.
Smart attendance tracking solutions have won particular interest from educators in the UK and Australia. For example, Victoria’s Department of Education resorted to facial recognition to monitor the whereabouts of students, letting teachers and staff access attendance data through a web dashboard or a mobile app.
The creators of LoopLearn, a facial recognition-based attendance tracking solution popular among Australian educators, claim that using facial recognition saves 2.5 hours of teacher time per week and helps effectively overcome the issues of fake attendance.
The core of the technology can be extended to enable more use cases, for instance, setting up automated book-lending library systems or managing other campus facilities.
Increasing learning engagement
Along with identifying and classifying faces, facial recognition systems can also interpret a wide range of emotions. Researchers report that analyzing students’ facial microexpressions, say, raising eyebrows or tightening eyelids, may help highlight boredom, confusion, delight, frustration, surprise, and other emotions. This can be useful for professors and curriculum designers. For example, when conducting a lecture, a professor may evaluate the emotional state of the attendees and determine the parts of the lecture that spark or weaken interest.
As the insights about student engagement come in, faculty may adjust the curriculum to better reflect student preferences and provide a more tailored learning experience. Still, today the applications of AI for emotion detection are primarily experimental.
Banking and finance
With financial services gone almost entirely digital, it is only natural for customer verification procedures to follow. Building on eKYC — a digital version of the “know your customer” standard that governs verification and authentication of a customer’s data — financial institutions can now shift the customer onboarding process entirely to online. While eKYC uses fingerprints as an authentication method, facial recognition may become a viable alternative.
Here’s an example of how the customer onboarding process may run when backed up by facial recognition. A customer requests to open an account. As a part of the account registration and eKYC, a bank clerk takes the client’s photo that is then linked to the client’s ID. Once digital authentication is complete, the client may let their face be scanned to access bank services in the future.
Running cardless ATM transactions
Today, criminals routinely use skimming devices to hack into ATMs. Facial recognition could potentially replace plastic cards and PINs as a more secure option for preventing fraud.
Several banks have already started testing out facial recognition solutions. For example, National Australia Bank partnered with Sydney-headquartered OCR Labs to develop a system that allows the bank’s customers to scan their faces in order to access ATM services. The solution captures images of the customers’ identity documents and matches them to the photos of customers taken by a camera integrated into the machine. The proof of concept was successfully implemented in 2020, and the bank continues to perfect the technology so that it meets the federal government’s data ethics framework.
The proponents of rolling out facial recognition in enterprise environments say the technology brings about many benefits. Here are the essential ones:
Although capable of driving tangible benefits, facial recognition is not flawless. Among the issues halting a wider adoption of the technology are the following:
If you consider rolling out an AI-based facial recognition solution, we recommend committing to doing so ethically.
First, at the very beginning of your project, you are likely to face a choice: choose a readily available solution, go the custom route, or opt for library-based development. To make the right decision, weigh the options against the objectives you intend to achieve — the more specific the task, the higher the need for custom software. If, on the other hand, you develop a facial recognition solution aimed at the general public, it may be a faster option to implement an off-the-shelf solution or API, for instance, Microsoft Face API or Amazon Rekognition, or build on an existing facial recognition library, for example, DeepFace, FaceNet, InsightFace, or others.
Another critical aspect is asking people for informed consent for collecting and storing biometric data, as well as for other purposes, like using one’s photos to train the algorithms further. While some facial recognition systems can de-identify the information, biometric data can hardly be wholly anonymized, so timely informing people is essential to maintain trust and transparency.
One more aspect to keep in mind is ensuring your solution is explainable. A user should understand why a system has come to a particular decision and revert it in case of false positives or false negatives. For example, rolling out a facial recognition solution, National Australia Bank intentionally chose to refer all user verification requests that the system could not verify to a human operator rather than rejecting them, which allowed reducing the error rate.
As of 2021, the facial recognition industry is still maturing. While facial recognition solutions are getting more affordable and easier to build, the promising potential of the tech is halted by incomprehensive legislation. Amid public debate about the safety of facial recognition technology, businesses and tech vendors should prioritize building transparent and explainable solutions.
If you still have questions about facial recognition technology or look for an AI technology solution boasting facial recognition, contact ITRex consultants. They will answer any of your questions.