Accidentally blinking in photographs may soon be a thing of the past, thanks to a new application of artificial intelligence developed by Facebook.
A team of Facebook (FB) engineers have successfully used a form of machine learning—known as generative adversarial network, or GAN—to automatically retouch closed eyes in photos, according to a research paper published by the company on Monday.
GAN uses reference images of the subject not blinking as a way of gleaning eye shape, skin tone, and iris color before applying that information to an image of a subject with closed eyes in a process known as in-painting, according to the paper.
GANs can also account for lighting conditions and poses in the in-painted photo, but the researchers admit that the process struggles with glasses and/or hair covering the eyes in either the reference images or the in-painted image.
It is not yet clear whether Facebook will release the technology for use on its social media platforms or apps, but researchers say that they may apply GANs, “to other in-painting tasks, such as filling in missing regions from a natural but uniquely identifiable scene,” in the near future.
