The State of Facial Recognition Software in 2019

Facial recognition (FR) is probably one of the most commonly accepted AI technologies in the category of image analysis software. People are used to things like Windows Hello and Apple Face ID. These are great examples of how FR is making our lives easier. Even pop culture is starting to take note of the tech, with stars like Taylor Swift using FR to identify unwanted concertgoers¹.

Its immediate convenience and the simplicity of the user experience ensure that companies will keep investing in facial recognition software to improve their products. That’s why more than 25% of AI startups with the largest investment rounds are focused on computer vision/FR².

At the same time, the quality of FR technologies has been steadily improving, paving the way to its much broader adoption across the globe. According to the research by the National Institute of Standards and Technology, the accuracy of FR has improved by 20 times in less than a decade³.

This report aims to shed some light on the broader state of FR technologies, their advances, and their much wider range of use cases, which some businesses and consumers might not be aware of. Additionally, we’ll take a look at some of the obstacles that FR tech has to overcome on its way to mass adoption.

You would worry if a stranger was walking next to you the whole day while simultaneously looking you in the eyes. Similar privacy concerns are associated with FR. Only this kind of intrusion raises a broader assortment of privacy and legality questions.

And these concerns are not baseless. In 2017, the UK police used FR to identify potential wrongdoers during the Champions League Finals, which led to over 2,000 people being wrongfully matched to actual criminals. In this particular instance, FR failed in 92% of the cases⁹.

Amazon, one of the leaders in AI and FR, also gets its fair share of flak for its products, like in 2018, when its system confused over two dozen members of Congress with criminals¹⁰. Some lawmakers even went as far as to call Amazon’s software racially biased¹¹. Sure, we might be inclined to insert a political joke here. However, it’s obvious that cases like these spark the general public’s concern about the future of FR.

Luckily, companies like ModiFace, recently purchased by L’Oreal, are already addressing possible privacy and bias concerns. They’re doing this by providing as diverse data to their facial recognition software as possible, with a curated database of 250,000 images that include people from most dominant ethnic groups¹². This way, when the system uses photos to learn how to recognize faces, it will be more trained to recognize racial nuances and better identify facial features.

Even schools are now experimenting with FR to monitor their students. For example, SAFR offers its FR software to K-12 schools for free¹³. This naturally alarms privacy rights groups and concerned parents¹⁴.

That’s why some of the biggest AI research organizations, like the AI Now Institute, are calling for regulations when it comes to FR tech¹⁵. Even Microsoft’s leading legal executive, Brad Smith, recently pointed out that FR software needs to be regulated¹⁶.

According to a prominent law firm, Drinker Biddle & Reath, companies adopting FR technologies have to be aware of the fact the regulations might soon catch up with the technology¹⁷. And this might require considerable investment and restructuring of the current processes. A great example of such a regulatory shift is the GDPR: FTSE 350 businesses spent over a billion dollars just preparing for it¹⁸.

Now, these circumstances should not dissuade any company from exploring the opportunities that come with FR—from improved security to better product personalization for customers. Let’s dive deeper into some of the use cases and industries that can benefit from FR in the nearest future.

The travel and hospitality industry relies heavily on the quality of customer experience to generate revenue. On top of that, customer experience is mostly shaped my interactions, like during a check-in, which creates opportunities and use cases for facial recognition software. This innovation is also greatly facilitated by the advances in FR tech, which make it easier to set up and operationalize new FR-based systems.

Augmented approaches to customer authorization in hotels and airports don’t always work since even mobile check-in becomes another friction point.

Facial recognition provides an opportunity to simplify customer authorization and streamline it even further. Marriott is rolling out its FR-powered check-in system at dozens of its properties²⁶. According to Marriott, this has the potential to shorten the check-in process by three times²⁷. Other hospitality businesses are also slowly rolling out similar systems, like the French Accor Hotels in one of its Brazilian properties²⁸.

Delta, one of the biggest US airlines, is also experimenting with FR for its international flights. This should improve the check-in experience and eliminate the already shrinking amount of paperwork required to get on a plane. Although the adoption rate still remains low, Delta is hoping to expand the impact by making FR available at more airports, such as Detroit Metropolitan Airport²⁹.

Federal authorities are also using facial recognition software to authorize passengers, as the U.S. Customs and Border Protection claims that these systems reach a 99% accuracy rate³⁰. The U.S. Customs and Border Protection has even caught an imposter with a fake ID with the help of facial recognition³¹.

There is a growing list of such systems working around the world to improve the check-in and customs experience for passengers. For example, Malaysia, France, and UAE are already exploring or installing facial recognition systems at their airports³². Beyond improving arrivals and departures of travelers and hotel guests, facial recognition can be used as a personalization technique:

• Instead of carrying a key card or identifying yourself to the hotel staff, a guest can get access to facilities and services based on their profile through the hotel’s facial recognition system.
• Hotels can match a guest’s face to the data stored in their database and suggest personalized offers based on the guest’s previous stays and available loyalty promotions.

Similar access control applications are being implemented in other adjacent industries that continually deal with large volumes of people. For example, products like Mexia One allow event organizers to abandon the legacy approach to ticketing and tagging by using facial data to grant access to their events.

It is, however, important to remember that 70% of consumers still prefer human interactions³³. Companies need to find the right balance between automating some of the above-mentioned functions and maintaining a human presence throughout the process.

At the moment, facial recognition is going through a cautious adoption curve that is affected by multiple privacy concerns across its application areas. Both businesses and governments are looking for ways to make sense of the ethical implications and balance them with the promising benefits of FR. Meanwhile, a few industries have successfully tested the waters, driving the FR implementations across retail, travel and hospitality, and healthcare industries.

Here are the key takeaways on the state of facial recognition software in 2019:

• The public sector is a major adopter of FR technology, especially in the U.S. and China.
• Racial bias and accuracy remain an issue yet to be figured out in emerging FR solutions.
• Retail applies FR to fight thefts and returns fraud as well as improve in-store experience.
• In travel and hospitality, FR is set to remove touchpoint frictions and make guest and traveler experience more enjoyable.
• Health care providers are bidding on FR to cut duplicate records costs, improve patient management, and make self-medication safer.