A new artificial intelligence imaging tool for user-interactive 3D manipulation of 2D images depicted in a photograph has been developed by a group of computer scientists from the Max Planck Institute for Informatics, MIT, Google, and the University of Pennsylvania. On the arXiv preprint server, the team published a paper and a few short videos demonstrating the new tool’s capabilities. The tool’s name is DragGAN.
Photoshop was first delivered in the last part of the 1980s, and since that time, it and comparative applications have been utilized to alter photos. Such use has turned into a standard piece of virtual entertainment — individuals photoshop pictures prior to posting them online as a way to “get to the next level.” By incorporating artificial intelligence into this new endeavor, the research team has elevated image editing to an entirely new level.
DragGAN resembles no other image manipulation tool at first glance. However, the creative team’s videos show that it can do things that no other application has ever been able to, like allowing users to change images on the fly in fictitious 3D. The findings are described by the researchers as “hallucinated occluded content.”
Photos, by their nature, are two-layered. Past photograph altering devices have considered obscuring, shading, or, in any event, fixing in other symbolism. However, the user must direct color correction or wrinkle blurring in all of this editing. With user prompting, an AI-based photo editing tool can infer what missing parts of a picture might look like by analyzing thousands or millions of other images and teaching it to recognize features.
For instance, with just a click and drag, a photo of an irate person in a video can be changed to show the same person smiling. The individual’s face can also be turned, uncovering portions of the head that were never caught in the first photo. Similarly, with just a few clicks and drags, cars, animals, or landscapes can be drastically altered. The addition of AI to the field of photo editing adds a whole new dimension to the field that has the potential to make as much of an impact as Photoshop did when it first came out.
More information: Xingang Pan et al, Drag Your GAN: Interactive Point-based Manipulation on the Generative Image Manifold, arXiv (2023). DOI: 10.48550/arxiv.2305.10973
