If you are the one who always bothers the photos in a group leave with closed eyes, you will be interested to know that Facebook has developed a technology that could be the solution to this problem . The system, developed after carrying out a, consists of “drawing” in a virtual way the eyes of the people that appear in the photograph, using machine learning systems to learn how to replace the closed eyes with others opened in a precise way and almost without being noticed.
The technology that uses this system is called Generative Adversarial Network , and consists, as I said, in an artificial intelligence that tries self-deceiving thinking that his creations are real. As they point, one part of the system is capable of recognizing faces, while another part is responsible for creating images that, thanks to the recognition algorithm, are increasingly realistic.
In the images on these lines, you can see how the system GAN Facebook – on the right – is able to generate much more accurate images to reality than the system included in Adobe Photoshop – on the left. This is because the technology developed by the company led by Mark Zuckerberg is trained with thousands of images , and therefore is able to reproduce photographs that fit the reality. The difference, if we take the reference image located on the left, is more than obvious.
As specified in the report, researchers from the social network decided to incorporate reference images of each person with their eyes open, so that the system could learn both the shape of the user’s eyes, such as color or style , as well as any characteristic feature. Thus, by providing images in which the eyes of the user in question remain closed, the system only has to demonstrate what it has learned and generate an image in which the eyes appear open and with specific features.
We do not know if this system will ever be a function incorporated in the social network , but it would certainly be an interesting way to take advantage of machine learning within the platform. Even so, the technology is still in development, because the samples obtained continue to fail in specific situations such as when the person’s eyes are covered by hair, or when the angle of the eyes is not the ideal one.