The neural network learned to identify the Photoshop on portraits

Science

Нейросеть научилась определять Photoshop на портретах

Adobe taught a neural network to determine whether they used Photoshop on portraits. So firm, in conjunction with the University of Berkeley is going to deal with fake photos, according to the website Theblog.adobe.com.

The algorithm is designed in such a way that the neural network not only determines corrected on photo space, but also tells you how to remove effects from the image. It operates with the help of Liquify tool, which is used in Photoshop to edit the face shape and the change of emotions.

Adobe say that they will not make the development of a commercial product. The experiment – trying to learn how to recognize manipulation of photos, video, audio and documents. Her task is to help professionals to deal with fake content. The algorithm is able to recognize a fake in 99% of cases, while people know only 53 % about the images.

The company said that they were thinking about creating a special “Cancel” button that would shut off all the effects. According to the developers, this is important, as is getting harder to trust digital information.

It is noted that this is the first such algorithm in the world. Adobe is also thinking to create a neural network to recognize changes in skin tone.

Share Button