Adobe uses machine learning to detect photos with Photoshop

Experts around the world are increasingly concerned that the use of AI tools will make editing photos and videos easier and easier - especially social networks with tremendous power with internal sharing capabilities. It is very fast and has not been verified for authenticity.

Adobe is developing a number of tools, and among them is a tool to help prevent this situation by using machine learning to automatically detect edited images.

Their latest product introduced at this month's CVPR conference shows that human-made digital forensics can be automated. Research is not a breakthrough in this area and has not become a commercial product, but it is also interesting to see Adobe - a name well known for photo editing software - create such a product.

The company's representative said 'the research project is still in the early stages' but in the future they want to contribute in 'developing technology to monitor and authenticate the authenticity of media products' .

Adobe uses machine learning to detect photos with Photoshop Picture 1
The famous photo of a rocket launch was launched in 2008 by the Iranian government

Specifically, everything is not clear because Adobe has never released a software to detect fake images. But they also mentioned cooperating with law enforcement units (such as using digital forensics to find lost children) as evidence for their responsible attitude towards the technology.

New research shows that machine learning can be used to detect three common photo editing styles: collage - two different images are merged into one, copied - objects in the image are copied and pasted to another image and deleted - The object is edited from other parts of the image.

To do this, digital forensic experts look for clues in image layers, often due to color disproportionate or brightness taken from the image sensor.

Adobe uses machine learning to detect photos with Photoshop Picture 2
The illustration in Adobe's new study shows the ability to detect edited images

Like many other machine learning systems, Adobe also uses edited image data sets. From here the AI ​​learns how to find common characteristics to detect what is edited. In some tests it is rated higher than other systems but not too much, nor capable of deepfake detection.

Anyway, there is a technology that detects things that are counterfeited in the digital world. AI can be a lot of worries but it can also be helpful.

See more:

1. Watching pictures painted by artificial intelligence, everyone thinks that is the work of a true artist
2. 6 steps to start learning artificial intelligence programming (AI)
3. How does Forensic work?