In a previous post, our computer vision consultants have described facial recognition algorithms on a technical level. In this post, we would like to take a more business-oriented approach and explore the real business value of this technology.
Each time you recognize someone’s face, you’re using an internal form of facial recognition. In a matter of milliseconds, your mind breaks down the parts of their face, puts them back together, and matches the sum with those faces already stored in your memory. When the process works seamlessly, you don’t even realize that it’s happening.
While you may not have given much thought to how your brain distinguishes one face from another, the behind-the-scenes process is fascinating and serves as the foundation for modern facial recognition apps. Though still considered an emerging technology, facial recognition is already being used in a number of applications ranging from social media to security. As more businesses consider applying this technology to their own organizations, it’s essential that they consider both the benefits and potential pitfalls of doing so.
How Does a Facial Recognition App Work?
We’ve covered this topic in detail in a previous blog post. Let’s briefly outline the basics necessary to understand the rest of this article.
Classified as a form of biometric identification, because it extracts information from the human body, much like retina scans and voice recognition, facial recognition software is designed to figure out what makes one face different from another. It does this by completing three key steps: face detection, faceprint creation, and identification. During this first step, the software picks out faces, as defined by the deep learning model it is based on. In short, the system is making sure that a face is a face and not, say, a ball or a pizza.
Once the system detects a face, it then begins making sense of it by creating a faceprint. Just as each person has a unique fingerprint, they also have a unique face. The distance between someone’s eyes, the length of their nose, and the shape of their mouth are just a few of the metrics that a facial recognition system uses to create a unique faceprint. While most systems can’t consistently differentiate between identical twins, some, like the iPhone X’s Face ID, are sophisticated enough to do so with a high degree of certainty. The result of all of these calculations is a mathematical formula that separates one face from another.
With the first two steps completed, the facial recognition software is now ready to identify faces. Some systems compare faceprints against the photo of a known person while others pull facial data from an existing database to help them identify an unknown person. In many cases, facial recognition data is simply cataloged and stored for future use.
Factors Affecting the Quality of Facial Recognition
The size, shape, and angle of a photo affect the ability of a facial recognition app to correctly identify faces in it. The image obtained in the initial stages may need to be resized or reoriented in order to improve the system’s recognition capabilities. Hi-resolution 2D and 3D images typically produce the best face matching results. However, one of the most significant factors influencing the quality of facial recognition is lighting. A photo with a well-lit front-facing face and unobstructed view is considered ideal. How far away a face is from the camera, the expression that the person is making in the photo, and the degree of contrast between the face and the background all have a similar impact on the efficacy of facial recognition systems.
How Businesses Are Currently Using Facial Recognition Apps
The applications for facial recognition are widespread. You’ve probably already interacted with this technology, perhaps without even realizing it.
Snapchat, the multimedia messaging app, is currently using facial recognition to help users set their own privacy setting for photographs. When a user sets their account to the specified setting, the facial recognition system scans the faces in the photo to determine which ones should be blocked out with an emoji to protect their privacy.
Android has a facial recognition app called Smart Lock, which allows smartphone owners to unlock their phones by holding it up to their faces. Apple is also working on a secure login facial recognition application.
Alibaba, the Chinese e-commerce powerhouse, is in the development phase of integrating its payment service, AliPay, with facial recognition software. This will allow customers to pay for their purchases without using sensitive payment methods like a credit or debit card. It will also help reduce the risk of fraud and theft.
Tesco, the UK’s largest grocery and general merchandise retailer, was among the first to add facial recognition software to its marketing strategy. In 2013, it announced that it would implement the technology at its petrol stations via OptimEye, which would then read customers’ faces at checkout to determine their age and gender. This information is now used to display tailored ads to patrons at checkout.
One of the earliest adopters of facial recognition technology, Facebook first started using it back in 2011. Anytime a user uploads a photo, the company’s facial recognition system systematically compares all of the faces that appear with that of the user’s friends. If a match is found, the interface suggests that the user tag their photo with the friend’s name.
The Potential Pitfalls
While most agree that facial recognition software has the power to revolutionize how businesses interact with consumers, there is also little doubt that in order for this technology to be successfully adopted en masse, the potential pitfalls must also be considered, and ideally, circumvented.
Checks and Balances
According to a report by the Georgetown Law Center on Privacy and Technology, the FBI has access to databases that contain facial recognition data for about half the US population. While the vast majority of this information is non-criminal in nature, the potential for misuse is very real. This applies to businesses as well, which is why organizations, big and small, need to make sure that they have the appropriate checks and balances in place before implementing facial recognition as part of their product or service offerings.
Anytime someone’s face is scanned by a facial recognition app, the results of that scan, specifically the mathematical formula that distinguishes that person from others, is stored somewhere in a database. Depending on who owns the database on which this information is stored, any number of third-parties may have access to it. Informing customers of how and when their information may be used (as GDPR demands) and obtaining their consent for such usage can go a long way towards establishing trust, in addition to preventing legal issues down the road.
Companies need to recognize that no technology, including facial recognition software, is infallible. Likewise, since facial recognition algorithms are trained using data collected by humans they are also not immune to bias. In fact, there have been several reported instances of facial recognition systems incorrectly identifying people with darker skin tones as the wrong gender or identified as criminals.
Facial recognition algorithms are only as good as the data they are trained on. The above scenarios occurred due to a lack of photos representing a diverse array of minorities and the over-representation of black men in mugshot photos. Companies can reduce these kinds of issues by making sure that facial recognition programs are properly trained with a sufficient and diverse amount of data.
With each new wave of technology, comes a new type of crime. The same facial recognition tools that allow the police to track criminals and find missing persons can be used to perpetrate crimes like stalking, theft, and fraud. Industrious criminals could access facial recognition data, either publicly or by hacking into a private database, to track persons without their permission. They would know when someone was home, at work or out of the country altogether, making theft significantly easier.
In addition, those with dubious intentions could also pretend to know people whose facial recognition data they have accessed, in the hopes of gaining sensitive personal information that could be used to commit fraud or even identity theft. In sum, the degree of damage that criminals would be able to inflict with the aid of facial recognition software would be notably greater. Knowing this can help companies prepare for this eventuality and provide them with a framework of information that could inform cybersecurity and facial data protection measures.
What Lies Ahead
Developments in the field of facial recognition are occurring at a rapid pace. While the widespread application of this technology holds much promise, it also needs to be handled with as much care. Businesses that want to use facial recognition apps need to understand this and approach their company strategy with care and consideration. Those that successfully manage to do this are sure to reap the benefits of facial recognition.