Now you see me, now you don’t— Sunglasses and Facial Recognition Technology
Let’s talk about faces. Most of us have them, and as facial recognition technology advances the concept of public anonymity will soon be antiquated. Depending on your personal beliefs, you may find this thought either comforting or frightening. But before we take a look at two different approaches to this divide, we must first look at the effect (if any) that glasses have on this type of technology.
First, there is a difference between facial detection and facial recognition, and the reality is that glasses are more likely to have an impact on facial detection.
Facial recognition is about determining from a picture who someone actually is. If you have a picture of a person, you want your software to be able to tell you with a high degree of certainty the identity of that person, e.g. “this picture is of Michael Jones, Employee Number 215”, or “the women in this picture are Stephanie and Jane”. Facial detection is the much more fundamental skill of determining whether there is a face in a picture at all. Obviously you can not have facial recognition unless you have already had facial detection.
People wearing glasses provide a challenge to both facial detection and facial recognition software. Glasses, especially reflective sunglasses, can hinder an algorithm from finding the points of reference it needs when determining whether there is a face in a photo. If there has been no facial detection, there will clearly be no facial recognition.
Every facial detection and/or recognition product uses a different algorithm. Many of these algorithms do recognize when people are wearing glasses. Prescription glasses with clear lenses will generally not be a problem for facial detection and recognition software because the key eye details are still visible. Even if they do cause some form of issue the software will continue to examine the other parts of the face and simply remove the occluded portions from its analysis. It is not unusual for faces to still be detected and recognized with up to 30% occlusion.
The best algorithms in the world still have difficulties, however, when people are wearing dark or shiny sunglasses, specifically ones that hide the pupils of the eyes. This problem is compounded when the glasses obscure or hide the distance between the eyes, which is a key part of many algorithms. Most facial recognition algorithms start by determining eye features first and comparing them against their pre-existing databases of images before they move onto other facial features.
Some companies believe that the future involves wearable tech that allows the user to wear sunglasses outfitted with augmented reality(AR) features that include facial recognition technology. One such product has been dubbed iFalcon Face Control Mobile by NNTC.
The AR glasses have an 8-megapixel camera embedded in the frame which allows the wearer to scan faces in a crowd and compare them with a database of over 1 million images. Notifications about positive identity matches are sent to the glasses’ see-through display, embedded in the lens.
NNTC claims that its facial recognition algorithms are in the top three for accuracy in the US government’s Face Recognition Vendor Test, able to detect up to 15 faces per frame per second, and capable of identifying an individual in less than a second.
NNTC says it’s so far produced 50 pairs of facial recognition-enabled glasses, and that they are “currently being deployed into several security operations” in Abu Dhabi, the capital of the United Arab Emirates. The company says the glasses are only on sale to security and law enforcement for now.
Police in China deployed similar tech last year, using the hardware at train stations to pick out suspects in a crowd.
And before you say that tech like this seems too futuristic or dystopian, it’s not functionally too dissimilar from what is already being used in the US and other Western countries. American Police can already use imagery collected from speed warning cameras, body cameras and CCTV cameras to search for suspects using facial recognition software.
There are, of course, numerous privacy and civil rights concerns associated with facial recognition. The algorithms that power this technology are prone to bias, and this can lead to false arrests and imprisonment. Some believe this gives police officers a new tool to discriminate against minorities. San Francisco has even gone so far as to ban the use of facial recognition, but the technology will undoubtedly continue to spread around the world.
So what if you prefer to live in anonymity, and you don’t live in San Francisco? Well, there are companies on your side. And if you’re ambitious, you can even handle it on your own.
For the stylish and tech savvy customer, Phantom & Ghost sunglasses block 3D facial mapping during both day and night. They keep your eye measurements from being analyzed by facial recognition on traditional security cameras using infrared. They can be designed as sunglasses or with prescription lenses. They block most forms of facial recognition without calling attention to the wearer. The designer and creator Scott Urban calls this newest line “privacy eyewear.” The glasses seek to, at least in part, stem the tide of facial-recognition technology and create an opportunity for customers to exist in society without being tracked.
The Phantoms are the latest iteration in privacy wear from Urban, whose 2016 Ghost frames (currently sold out) reflect both visible light and infrared light as a way to partially blind traditional flash photography and IR security cameras.
Unlike the Ghost model, the newer Phantoms reflect only IR light and also come with the option of IR-light blocking lenses. This means that your face won’t be blurred out in photographs taken with a flash, or in standard security camera footage in a well-lit environment.
The frames are specifically designed to defeat 3D dot matrix face-mapping systems, which is basically what makes Apple’s Face ID work. They bounce infrared light back at its source, with the goal of preventing IR video cameras from getting a good image of your face — or potentially even registering your face as a face at all.
And for the DIY’er, there’s an option for you too. Researchers at Carnegie Mellon University developed a 3D-printed pair of sunglasses that only cost $.22 to manufacture. The sunglasses, according to an accompanying study, fools advanced facial recognition software by altering small bits of color information in a face.
Facial Recognition software relies on neural networks to compare pixel coloration in specific areas of the face to match one image to another. When something (like these glasses) alters this pixel coloration, it confuses the software. Researchers say the 3D printed glasses are good at both impersonation and dodging attacks. Impersonation “seeks to have a face recognized as a specific other face,” whereas dodging is more of an attempt to avoid being detected at all. Overall, the system achieved a 90 percent success rate in foiling facial recognition software.
I think we can agree that facial recognition technology is here for the long haul. As a junior developer, I enjoy watching the growth on both sides of the public safety vs. personal privacy debate as humans push each other to devise new and innovative ways to meet the demands of our changing world.
