Adobe is developing an artificial intelligence that can recognize photos changed by Photoshop. The company is concerned about combating fake content, created or not on its image editing platforms.
The example of the following image, made by Adobe, shows us the logic of Artificial Intelligence. The original picture is analyzed and changes are detected immediately. The AI concludes thus that it is an altered photograph.
In fact, it would be expected that an artificial intelligence developed by Adobe can detect the changes made by Photoshop, which also belongs to Adobe.
Artificial Intelligence was able to detect changes rather than humans
During the tests, Adobe asked a group of volunteer people to analyze images, looking for changes. The results were median for the volunteers where 53% of the changes were detected correctly.
However, Artificial Intelligence was able to detect altered images with an incredible accuracy of 99%. AI can then suggest ways to restore the photo to its original format, with some flaws.
This tool can help you detect Deepfakes
Deepfake content consists of images or videos where the faces of real people are manipulated to do or say what the creator of the content wants. The following video shows an excellent example where actor Jordan Peele talks over a fake video of Barack Obama, imitating his voice.
Deepfake videos are made by first filming the original person, with the desired expressions or phrases. Next, software is used to detect facial movements and place them in the face of the “victim” of Deepfake.
This content has become extremely popular initially as a way of putting celebrity faces into adult content. Therefore, the tool that Adobe is developing may be important in detecting these images.