Google and Facebook are building large Artificial Neural Networks (ANN) that can instantly recognize faces, cars, buildings, and other objects in digital photos. But now they are using Artificial Neural Networks for creating artistic images.
Facebook aims at creating realistic images using two techniques to generate tiny thumbnail images. Facebook’s technique is quite similar to the random vector which is same as how people used to learn painting from the scratch. The first algorithm creates a picture using the random vector and the second algorithm checks whether the images look realistic or not. If the image created by the first algorithm is not realistic, then it will be rejected automatically. Facebook research states “Around 40 percent of the samples generated by our class conditional LAPGAN model are realistic enough to fool a human into thinking they are real images.”
In comparison to Facebook, Google does not aim for realism. Instead, it’s producing images by letting the neural network run wild and decide on the visual elements that it wants to emphasize. Google wrote in a Blog Post explaining the project “This creates a feedback loop: if a cloud looks a little bit like a bird, the network will make it look more like a bird. This in turn will make the network recognize the bird even more strongly on the next pass and so forth, until a highly detailed bird appears, seemingly out of nowhere.” Google’s system tends to generate more abstract images and has named this technique as “Inceptionism”.
Both Google and Facebook are relentlessly working on to take AI one step-further by improving Artificial Neural Networks. In future we can hope to see machine-generated images to become more realistic and lose their trippy character.