Deep Forger: Art Forgery Meets Deep Neural Nets

The past year has seen deep learning make exceptional advances in imaging, perhaps most notably with Google's Deep Dream. See how a clever Twitter bot employs deep neural nets to paint images in the style of famous painters.



First, I submitted my profile pic, attached to the following simple tweet:

  @DeepForger #portrait +Monet


What I got back is shown below. For what it's worth, Deep Forger seems to think it could fetch $95K on the open market.

Deep Forger Example
Me, as a Monet.

We have now seen a few examples of how Deep Forger works on portraits and people, and so I thought it would be fun to try it out on some other image types. And what better test of a deep learning technology's artistic savvy than to have it render the KDD Conference 2011 poster as an abstract painting?

  @DeepForger /abstract


Deep Forger Example
Jackson Pollock's take on KDD.

Deep Forger chose Jackson Pollock as the 'semi-random' painter's style to use for the request. It makes me look at KDD in a different light, to be honest.

And how about this classic image of a data scientist solving a rather intricate problem, as would have been envisioned by Picasso?

  @DeepForger +Picasso


Deep Forger Example
Abstracting this problem does not make it seem any easier.

And finally, here is a generic AI brain image, painted in the style of Munch, which Deep Forger seems to think could command upwards of $720K.

  @DeepForger +Munch


Deep Forger Example
Munch's impressive (and valuable) take on the AI brain.

As is visible with this small sample size, Deep Forger (and its underlying technology) does better with some combinations of images and styles than others. Beauty is definitely in the eye of the beholder, however, and so one man's (woman's) neural network-produced trash is another's forged beauty.

I will say this: the more I look at and appreciate Munch's AI brain forgery, the more I think he may have been born a few years too soon.

Bio: Matthew Mayo is a computer science graduate student currently working on his thesis parallelizing machine learning algorithms. He is also a student of data mining, a data enthusiast, and an aspiring machine learning scientist.

Related: