Deep, or machine learning is a process of “teaching” a computer to accomplish a task by giving it vague outlines and letting it fail until it can figure out a problem.

We’ve written about it being used to teach RC cars to drift and master complicated games, but now it’s being used to create art.

The process, which is outlined in a scientific journal titled “Combining Markov Random Fields and Convolutional Neural Networks for Image Synthesis” looks at two images: a base image and a “style” image.

Using a process similar to how neural networks work in the brain, a computer can “learn” the distinct artistic look of the style image, and then apply that to the base image. This means that famous painters and artists long dead can be used to create new art.

Neural-Networks-Paint2

While all the above examples of deep learning may seem superficial, they are all steps towards creating artificial intelligence which can learn like human beings.

While that’s all fine and well, we bet you want to create some of your own art. Luckily, the complete source code for this project is available on  GitHub. All you’ll need to do is grab the source code and compile it.

To make it even simpler for us, some enterprising individuals have already created sites to do the work for you.

We’ve been pointed to two sites that you can use. Dreamscope is the simpler of the two, with a web interface and downloadable app. This one cheats a bit by turning certain styles into filters which are applied to images.

On the other hand, Deepart looks to be the real McCoy. The only drawback here is a massive wait to get your image. We gave the site something to work with and it said it would email us the result in 130 hours. You could wait the time out, or pay to have it rushed. We suggest waiting a few days for the servers to be less busy.

Make your own:

 

[Source – Reddit , Image – Imgur]