While ugly selfies are just part of our every day life, Facebook wants to eliminate that problem. The platform is working on fixing images where you blinked with a new tool that supposedly produces very realistic results. Not creepy at all.
-
Related Story: How Much Sex, Violence And Hate Speech Can Facebook Control?
Two engineers from Facebook published a paper detailing how the technology works, explaining that it uses a machine learning method known as generative adversarial network (GAN). This method has become sort of popular lately due to the realistic images it creates. In order to successfully recreate a photo, the software has to learn how the users’ eyes normally look by training it on images where the person is not blinking. After this, the software can fill in the images where people are blinking, even though glasses, extreme angles and other obstruction make the software work less efficiently.
There’s no date of release but it seems like this feature is a perfect fit for Facebook, who already has a vast catalogue of images of you to train its software on.
While this technology is similar to what other photo editing softwares are doing, it demonstrates how easy it is to create fake images when you have all the correct information. According to The Verge, this technology will reach the mainstream soon since Adobe and Pixelmator are also working on similar AI tweaks. Editing out blinking eyes is pretty innocent in the grand scheme of things, but the more features like this develop, the more the line between real and fake will continue to blur.
