Adobe probes tool that detects, reverses facial digital manipulation to fight deepfakes
Adobe, Twitter, and the New York Times have recently joined forces to draw attention to image manipulation and come up with an industry standard to fight deepfakes by identifying who is behind an image or video and whether they were altered in any way, writes Axios.
On this note, during its Max conference in Los Angeles, Adobe presented a new experimental feature that instantly tells if an image was digitally manipulated and lets you undo the edits, writes The Next Web. Called ‘About Face,’ the tool lets you upload an image and runs a detection algorithm on the pixels to create a heatmap of all the areas on the face that were photoshopped with the liquify feature.
The projects showcased at this event are often included in Adobe’s product portfolio. Adobe could introduce an authentication system by letting content creators link attribution data to their work, but more will be discussed at the company’s summit next month.
The report notes that most early attempts to deal with deepfakes have relied on blockchain to provide unalterable, decentralized records, but these may be difficult to interact with. The alternative is a database held by a single party.
“When it comes to the problem of deepfakes, we think the answer really is around ‘knowledge is power’ and transparency,” said Adobe general counsel Dana Rao. “We feel if we give people information about who and what to trust, we think they will have the ability to make good choices.”
Truepic, a startup that develops photo and video verification, has also joined the fight against deepfakes and believes that proving authenticity could be the way to do that.
Twitter recently partnered with Amazon to fight deepfakes, which have grown in popularity on social media.