Some Instagram photos are about to lead to some awkward conversations.
It’s getting harder and harder these days to know what somebody really looks like, with the rampant use of filters and photo editing software. But for people who tend to depend on such applications in order to alter their facial features, it seems like their days are now numbered.
Adobe, in collaboration with UC Berkeley researchers, has released a new experimental AI tool that can detect whether someone’s face has been photoshopped or not.
To test the AI, the researchers created several images using Photoshop’s “Face Aware Liquify” tool and mixed them with real edited photographs. The researchers then showed both types of images to human participants, who spotted the fakes 53% of the time. The AI, however, spotted the fakes with a whopping 99% accuracy.
“It might sound impossible because there are so many variations of facial geometry possible,” says one of the researchers, UC Berkeley professor Alexei A. Efros. “But, in this case, because deep learning can look at a combination of low-level image data, such as warping artifacts, as well as higher level cues such as layout, it seems to work.”