Adobe built an AI to keep Photoshopped faces honest

Yolanda Curtis
June 17, 2019

The company says it doesn't have any immediate plans to turn this latest work into a commercial product, but a spokesperson told The Verge it was just one of many "efforts across Adobe to better detect image, video, audio and document manipulations".

But with the threat of image and video fakery quickly reaching epic proportions (think deepfakes and face swapping), fake content, images and videos are becoming a major issue.

The team working on the project trained a convolutional neural network (CNN) to spot changes in images made with Photoshop's Face Away Liquify feature, which was created to change people's eyes, mouth and other facial features. It consists of training a Convolutional Neural Network (CNN), which is a form of deep learning, with a set of images.

The researchers proudly said that the AI is better than people at detecting edited from non-edited images. The company's researchers are working with UC Berkeley scientists and DARPA, the Pentagon's advanced technology research arm, to develop software that is capable of flagging, analyzing, and even reversing facial manipulation in photographs. The researches also informed that this is a step to a safer Digital Media and can restore people's trust towards online pictures and images.

As part of the programme, the team trained a convolutional neural network (CNN) to spot changes in images made with Photoshop's "Face Away Liquify" feature, which was intentionally created to change facial features like eyes and mouth. It allowed the user to revert back to the original state of the edited image. In addition, an artist was hired to alter images that were mixed in to the data set.

Adobe's execution in identifying facial control came only days after doctored videos of Facebook CEO Mark Zuckerberg and US Speaker Nancy Pelosi made the rounds via social media just as news channels. It was trained by showing pairs of images, one original and one altered, so the system could learn the telltale signs of manipulation.

"It might sound impossible because there are so many variations of facial geometry possible", said Professor Alexei A. Efros of UC Berkeley.

"The idea of a magic universal "undo" button to revert image edits is still far from reality".

Their tool was able to spot altered faces 99% of the time, in comparison to the human eye, which found the alterations 53% of the time, Adobe said.

Other reports by iNewsToday