A joint project between Adobe and UC Berkeley researchers, led to the development of an AI (artificial intelligence) tool that can detect images edited with Photoshop’s Liquify feature, which can be helpful in image forensics, the science of analyzing changes to digital images.
While Photoshop's Face Aware Liquify tool offers set of human-doctored photographs, whereby the original image and the doctored version seem undifferentiated to human inspection with spotted fakes only successful 53 per cent of the time, but with the Adobe AI it easily hit a 99 per cent accuracy.
The researchers trained a Convolutional Neural Network (CNN) on a set of modified images using the Face Liquify feature of Photoshop and artists were hired to train the AI for human-made alterations and techniques. Adobe is aiming to make it easier to spot a fake, thereby making the web world safer from misinformation as the issue of fake news has been blown out of proportion with the aid of professionally edited pictures becoming harder to smoke out.
It is coming as a part of a broader effort to better detect image, video, audio and document manipulations by Adobe. With past research haven focused on image manipulation detection from splicing, cloning, and removal; this present effort focuses on the use of Face Aware Liquify feature in Photoshop as it’s the most popular for adjusting facial features, including making adjustments to facial expressions.
The effects can be delicate which made it an intriguing test case for detecting both drastic and subtle alterations to faces, according to Adobe. Albeit, a magical ‘undo’ button won't turn up overnight, but the AI tool is perhaps a major step forward in spotting doctored images that spurs the spread of fake news.