This New AI-Driven Tool Lets You Edit, Remove, And Add Objects To Pictures With Just One Click

In the last few years, it is undeniable that software used to edit photos has evolved a great deal. Now, with the capabilities of new AI, it’s way easier to customize pictures the way you want. Researchers from IBM and MIT have developed a new AI that allows users to remove, edit, or add an object just by one click.

The tool GAN or Generative Adversarial Network Paint Studio allows users to upload pictures and customize them without diminishing any original details. For instance, if you want to add grass or trees to the existing photo, related objects will automatically be rectified. In that way, the final picture would become more realistic.

Take a look at the photo below. Some customization has been made, and it still looks quite natural.

An MIT’s professor – Antonio Torralba said that the GAN tool’s personality is strong and that it would refuse to add details if it believed that they did not fit in the picture. He also shared that although most of the drawing apps would follow user instructions, GAN would do things a bit differently. If users order it to put objects in impossible locations, the tool will refuse to do it.

The tool was initially tested by removing particular objects from the scene. Then, researchers observed how artifacts around these objects were affected. After that, they used this data to teach the algorithm so that it can make realistic modifications to photos once objects are removed or added.

The developers fully understand that GAN can be used for bad causes. Nevertheless, Jun-Yan Zhu, the co-author of the paper strongly believes that the tool is capable of helping researchers to understand mistakes that AI often makes and therefore decrease the number of fake images.