New technology from Facebook engineers might help you stop regretting those photos where your eyes are closed.
Two Facebook engineers published a new paper detailing a new method that uses machine learning to fix your eyes when theyre closed in photos.
The researchers said the technology could open your eyes using a similar method as removing red-eye or smoothing out blemishes.
According to The Verge, the technology uses the generative adversarial network, which has previously been used to create fake celeb photos, change how weather appears in videos and design clothes.
Blinking is no challenge at all, The Verge reported.
The system would start by learning what your face and eyes look like on a normal day when you're not blinking.
Then, when you take a photo in which you are blinking, the tech can rework the picture to show your eyes wide open.
Its not a perfect solution, and the AI struggles to edit glasses, long fringes, and faces at extreme angles. But when it works, it works; creating incredibly realistic fakes, according to The Verge.
However, according to The Sun, Facebook didnt succeed in its first attempts. Many of the initial photos attempting to correct blinks werent convincing enough.
Thats when the researchers realized the tech needed to use example photos of users so technology would understand what the person looks like on a normal day.
The company hasnt unveiled how it plans to use the technology.
Last December, Facebook announced it would add more facial recognition software to help notify you when someone adds a photo of you on Facebook, even if youre not tagged, according to Wired.
"They wont identify you using face recognition to people who couldnt identity you in real life, and that to me seems like the right line," Chris Calabrese, vice president of policy at the Center for Democracy & Technology, told Wired. "I personally am comfortable with face tagging in this very circumscribed context, but only in that context where its to someone who would already recognize you. If we cross that line, face recognition could rapidly spin out of control, and that could be really problematic."
Two Facebook engineers published a new paper detailing a new method that uses machine learning to fix your eyes when theyre closed in photos.
The researchers said the technology could open your eyes using a similar method as removing red-eye or smoothing out blemishes.
According to The Verge, the technology uses the generative adversarial network, which has previously been used to create fake celeb photos, change how weather appears in videos and design clothes.
Blinking is no challenge at all, The Verge reported.
The system would start by learning what your face and eyes look like on a normal day when you're not blinking.
Then, when you take a photo in which you are blinking, the tech can rework the picture to show your eyes wide open.
Its not a perfect solution, and the AI struggles to edit glasses, long fringes, and faces at extreme angles. But when it works, it works; creating incredibly realistic fakes, according to The Verge.
However, according to The Sun, Facebook didnt succeed in its first attempts. Many of the initial photos attempting to correct blinks werent convincing enough.
Thats when the researchers realized the tech needed to use example photos of users so technology would understand what the person looks like on a normal day.
The company hasnt unveiled how it plans to use the technology.
Last December, Facebook announced it would add more facial recognition software to help notify you when someone adds a photo of you on Facebook, even if youre not tagged, according to Wired.
"They wont identify you using face recognition to people who couldnt identity you in real life, and that to me seems like the right line," Chris Calabrese, vice president of policy at the Center for Democracy & Technology, told Wired. "I personally am comfortable with face tagging in this very circumscribed context, but only in that context where its to someone who would already recognize you. If we cross that line, face recognition could rapidly spin out of control, and that could be really problematic."
