Generative adversarial networks (GANs) — two-part neural networks consisting of generators that produce samples and discriminators that attempt to distinguish between the generated samples and real-world samples — have been used to discover new drugs, create convincing photos of burgers and butterflies, and generate synthetic scans of brain cancer. And as a new paper published by Maastricht University in the Netherlands reveals, they’re not half bad at generating logos, either.
In research published on the preprint server Arxiv.org (“LoGAN: Generating Logos with a Generative Adversarial Neural Network Conditioned on color“), scientists describe an artificially intelligent (AI) system — dubbed LoGAN — that can craft logos from 12 different colors.
“Designing a logo is a long, complicated, and expensive process for any designer. However, recent advancements in generative algorithms provide models that could offer a possible solution,” they wrote. “LoGAN’s results offer a first glance at how artificial intelligence can be used to assist designers in their creative process and open promising future directions, such as including more descriptive labels which will provide a more exhaustive and easy-to-use system.”
The problem with art-generating GANs, the researchers explained, is that they don’t always produce aesthetically pleasing results. Their solution was to define logos using their most prominent color: black, blue, brown, cyan, gray, green, orange, pink, purple, red, white, and yellow.
The team trained the system on the LDD-icons dataset, which consists of 486,777 small icons measuring 32 by 32 pixels. The dominant color in each was extracted with an algorithm and converted from the RGB values into color words. Simultaneously, a third neural network in the system (in addition to the generator and discriminator) classified the sample images.
So how’d LoGAN do? Despite the generated logos’ blurriness (due to the source images’ low resolution), some of the logos were pretty convincing. When fed color keywords, it managed to come up with irregular shapes, round and square buttons, and even a lookalike of the Google Chrome logo.
White and grey, interestingly, were among the top three most common color combinations across the 12 color classes. In the orange class, brown was a go-to for the neural net, and in the yellow class, it commonly picked blue.
The researchers believe AI systems like LoGAN could handle some of logo design’s more monotonous work, freeing up designers to brainstorm. In future work, they hope to expand the system’s semantic understanding of words to keywords beyond colors, such as shapes and focus.
An improved system might be trained on two distinct datasets, they wrote: one containing logos with an obvious geometric shape, and a second with non-regular shapes. And it might use an embedding model with most-used words describing the logos to “boost interpretability.”
“[While] the generated logos have very low resolution, they can serve as a very rough first draft of a final logo, or as a means of inspiration for the designer,” the researchers wrote. “The proposed model can successfully create logos if given a certain keyword, which in our case consisted of the most prominent color in the logo. This class of keywords can be considered descriptive as it provides a property of the logo that is easy for humans to distinguish.”
Leveraging the power of AI to produce artwork isn’t a new idea, it’s worth noting. Botnik Studios, a graduate of Amazon’s Alexa Accelerator program, recently taught a neural network to write a satirical Coachella poster with a list of fictional band names. Prisma, a popular smartphone app, uses a machine learning technique known as style transfer to make photographs appear as though they’ve been executed in paint. And game design AI startup Promethean AI automates the process of building out virtual landscapes and interiors.