Facebook develops AI to open your damn eyes in photos

Karen, you don't need a puppy dog filter, you just need to stop blinking in the Christmas photo!

Claire Reilly Former Principal Video Producer
Claire Reilly was a video host, journalist and producer covering all things space, futurism, science and culture. Whether she's covering breaking news, explaining complex science topics or exploring the weirder sides of tech culture, Claire gets to the heart of why technology matters to everyone. She's been a regular commentator on broadcast news, and in her spare time, she's a cabaret enthusiast, Simpsons aficionado and closet country music lover. She originally hails from Sydney but now calls San Francisco home.
Expertise Space | Futurism | Robotics | Tech Culture | Science and Sci-Tech Credentials
  • Webby Award Winner (Best Video Host, 2021), Webby Nominee (Podcasts, 2021), Gold Telly (Documentary Series, 2021), Silver Telly (Video Writing, 2021), W3 Award (Best Host, 2020), Australian IT Journalism Awards (Best Journalist, Best News Journalist 2017)
Claire Reilly
2 min read

Facebook wants to recreate eyes in photos using AI. 


There's always one person. Seventeen takes. And Karen's blinking.

But no need to host an intervention -- Facebook plans to fix your photos with AI.

The company released a research paper this week, authored by two Facebook engineers, which outlines a new method of recreating eyes in photos using what's known as a "exemplar generative adversarial network" or ExGAN. It's the kind of deep learning that's already used to generate pictures of faces from scratch and now Facebook wants to use it to fix your selfies.

According to the researchers, regular deep neural networks can learn to recreate eyes in photos, but there's a way to make them look like that person's real eyes. Put another way, your Karens could end up looking like Darrens.

Watch this: Facebook fixer-upper: Can artificial intelligence clean up your feed?

Rather than teaching a deep network to recreate eyes from a data set of other people's eyes, the researchers figure the AI can use photos of the same person as a reference point. And that's where Facebook has the upper hand -- it already has countless photos of you tagged on its servers, ready to cross-reference.

Facebook is already using machine learning to help identify faces on its platform and to give you tagging suggestions. And while the current technology is just in the research phase, it's a sign that the social network wants to maintain its dominance as a photo sharing platform. With more than 350 million photos uploaded to Facebook, the company wants to make sure everyone is a keeper.

Follow the Money: This is how digital cash is changing the way we save, shop and work.

CNET Magazine: Check out a sample of the stories in CNET's newsstand edition.