Adobe machine learning model must detect certain types of custom photos




Adobe, known for products such as Photoshop, has presented artificial intelligence research to detect certain types of manipulated images. With this, the company claims to want to contribute to trust in digital media.

Adobe writes based on previously published research that its machine learning model should be able to detect three types of modifications. Firstly, that is splicing , in which elements from two images are combined. The second technique is copy-move , which means that an element from the same image is moved or copied. Finally, the removal and subsequent filling in of elements from an image must also be possible.


Adobe’s paper claims that combining two input streams, consisting of a rgb stream and a noise -stream, yields better results than other methods for analyzing manipulated images. Using the rgb stream, for example, it is possible to determine whether there is unusually high contrast at the edges of an object in an image. The second stream is then useful to determine whether a particular element comes from another image.

The company tells The Verge that his research is still in its infancy. It adds that it ‘wants to play a role in the development of technology that supports the tracking and verification of the authenticity of digital media’. It is therefore unclear whether the research indicates an Adobe product in this direction. So far, the company has not yet released such a product in addition to its image processing software. Tweakers wrote an article about the feasibility of image and sound .

adobe ml1


In: A Technology & Gadgets Asked By: [23167 Red Star Level]

Answer this Question

You must be Logged In to post an Answer.

Not a member yet? Sign Up Now »