Since last year 2020, masks have been present in many western countries due to the pandemic. They offer security to citizens against the pandemic, but they are a problem for facial recognition systems, since they hide various characteristics of the face necessary for this software to be able to identify people. But like almost everything in technology, ending this problem was only a matter of time. And it will stop being an obstacle, thanks to Clearview AI, a company that according to IT Business Edge is dedicated to the development of facial recognition software for law enforcement agencies, which has a photo database of more than 10 billion images and which claims to have solved the problem.
Facial recognition uses Artificial Intelligence to analyze the geometry of a person’s face, such as the distance between the eyes or the shape of their chin. But until now, if many of the facial features are covered, these systems do not work. What Clearview AI has done is design a pack of mask removal tools that you would use the images you have in your database with Artificial Intelligence to guess what face the person hiding under the mask may have.
The model takes data points from the part of the face that it can see, such as the eyes, forehead, or ears, and tries to square them with the images it has. To do this, the software uses statistical patterns in charge of determining the facial features that are hidden. These features can be useful for publicity and emotion recognition, as long as organizations use them with permission. Mind you, many of Clearview’s clients are law enforcement agencies.
Although the company claims to have solved the problem, what its software really does is guess what the face of the person you want to identify would be like, and guessing can lead to serious errors. And since facial recognition has problems with exposed faces, especially in the identification of women of color, having to guess what two-thirds of a face are like makes the situation more complicated, something to take into account. And just many of its users, especially security agencies, do not seem to realize, or face, that this technology is very inaccurate.
In addition, as has recently become known, Artificial Intelligence algorithms require human supervision to achieve the best possible results, something that under-staffed agencies may not do. Especially if they do not see any benefit. Hoan Ton-That, one of the founders of Clearview AI, and its CEO, points out that his intention with this technology «it is to always have it under human control. When Artificial Intelligence is wrong, it is checked by a person«.
Artificial Intelligence facial recognition is not yet accurate enough to lead to life-changing decisions. But companies can use it to improve their product lines or give their employees access to devices without password. Using this technology in these ways avoids the problems associated with privacy in facial recognition, and gives you a chance to improve its accuracy.
