Pretty soon artificial intelligence in photography is going to be so advanced you probably won’t even need to show up for the shoot in the first place. Heck, it might not even require a photographer or being on location in general at the rate things are advancing.
In yet another advance in the field of AI and digital imagery, Facebook has released an advance that cures one of the most common problems to afflict family photos and yearbook pictures all over the world – the shut eye phenomenon.
It has happened to all of us. An otherwise perfect group perfect is thrown off by one person with their eyes shut or that ideal selfie, you know the type, is ruined by a meditative look on your face rather than an open-eyed smile.
Facebook’s new development uses machine learning to basically generate what the photo would look like had you not ruined everything by shutting your eyes. Brian Dolhansky and Cristian Canton Ferrer detailed their research in a publication from Facebook that described how the AI manipulating the photos did so through what is called a generative adversarial network or GAN. The AI uses a technique the researchers called “in-painting” to fill in the missing details using other photos wherein your eyes are open. This means that the AI-generated recreation of your open eyes will match in terms of color and look. While not perfect, some of the examples on offer from the research team are quite striking.
But don’t worry – the research team is already ahead of us in terms of figuring out whether or not people can tell the difference. When asked to evaluate the photographs, people couldn’t tell the difference 54% of the time. The technology is being refined to include users with eye glasses as well as those who may have some kind of obstruction blocking a full view of their eyes such as a strand of hair.