Facebook uses smart techniques to make it more and more interesting for people to share photos. With artificial intelligence (AI), even photographs can become post-worthy once someone has their eyes closed on the photo. How does this work exactly?
Facebook developers are investigating how machine learning can be applied to update closed eyes on photos. Removing red-eye and other unevenness has been possible for a long time with different tools, so why not also dense eyes, the developers wondered. The research is still underway, but it is not excluded that Facebook will ever introduce this feature.
The editing of closed eyes is possible by a machine learning method called GAN (generative adversarial network). For example, this system can already make fake celebrities, adjust the weather in videos and design clothes. Blinking should not be a challenge.
First, the system must know what someone’s eyes normally look like. The system is trained, as it were, to learn how open eyes look. This way it learns how the shape and color of eyes is. If someone blinks on a photo, and the eyes are closed, the system can ‘fill in’ the open eyes on the basis of what it has learned.
The application is not optimal yet, because glasses and faces with a sharp angle is a tricky story. But it does work, as can be seen in the image below.
Editing the future of photos
In addition to Facebook, more and more photo programs such as Adobe and Pixelmator are also working with AI. The possibilities are becoming more and more extensive and it is even conceivable that completely new photos can be made with the help of AI.
A question is whether this development contributes to natural photographs. After all, we are not perfect, right? That wrinkle and your freckles may be seen. Although it is nice to be in the picture with open eyes, so, for now, we thank the developers of Facebook for a while.