New AI research lets you click and drag images to manipulate them in seconds

No, it’s not over yet: the ability of AI tools to manipulate images continues to grow. The latest example is only a research paper for now, but a very impressive one, which lets users simply drag elements of a picture to change their appearance.

It doesn’t sound very exciting, but take a look at the examples below to get an idea of ​​what this system can do.

Not only can you change the dimensions of a car or turn a smile into a frown with a simple click and drag, but you can also rotate the subject of a picture as if it were a 3D model – for example Change the direction of facing. A demo also shows that the user adjusts the reflections to the height of a lake and a mountain range with a few clicks.

Here is an overview on the different topics:

Here’s a closer look at landscape manipulation:

And just for fun, messing around with lions:

these come from the video Research Team HomepageHowever it is crashing due to the amount of traffic Twitter sends to the site (Mainly by user @_akhaliq, which does a great job of highlighting interesting AI papers and is worth following if you’re interested). You can also read research papers on arXiv right here,

As the team responsible notes, what’s really interesting about this work isn’t necessarily the image-manipulation. In fact, but the user interface. We’ve been able to use AI tools like GANs to generate realistic images for a while, but most methods lack flexibility and accuracy. You can ask an AI image generator to “make a picture of a lion chasing you through the savannah”, and you’ll get one, but it might not precise currency you want or need.

This model, called DragGain, offers a clear solution to this. The interface is very similar to traditional image-warping, but instead of simply blurring and crushing existing pixels, the model generates subjects anew. As the researchers write: “(o) your perspective can confuse occluded material, such as the teeth inside a lion’s mouth, and can be distorted following the object’s rigidity, such as the hoof of a horse’s leg.”

Obviously this is just a demo for now, and it’s impossible to fully evaluate the technology. (For example, how real are the final images? It’s hard to tell based on the low-resolution video available.) But this is another example of image manipulation being made more accessible.