2 min read Data Science

Making Art with Generative Adversarial Networks

Can computers make art? To find out, we tried ourselves. We used Generative Adversarial Networks to try to paint new Van Gogh paintings.
Making Art with Generative Adversarial Networks
Note: the people on the left are not real people. Nor is the art next to it. They were created by an AI – to be specific, by a Generative Adversarial Network.

Generative Adversarial Networks (GAN's) are a relatively new type of technique for generating samples from a learned distribution, in which two networks are simultaneously trained whilst competing against each other. Applications for GAN’s are numerous, including image up-sampling, image generation, and the recently quite popular Deep Fakes. In this project, we aim to train such a Generative Adversarial Network ourselves, with the purpose of image generation, specifically. As the generation of human faces has been widely studied, we have chosen a different topic, namely: the generation of paintings. While large datasets of paintings are available, we have opted to restrict ourselves to one artist, as we believe this will give a better chance at producing realistic paintings. For this, we have chosen the Dutch artist Vincent van Gogh, who is known for his unique style.

How

There are many GAN architectures around. Some popular of which the DCGAN and the StyleGAN. We decided to train both and compare the results.

  • DCGAN. A GAN architecture that has been around for a while. Was trained using TensorFlow.
  • StyleGAN. A popular GAN architecture for generating faces, provisioned by NVIDIA Research. Also trained using TensorFlow.

Both were trained on the Van Gogh dataset, as available on Kaggle. Because the dataset contained only a limited amount of paintings (painting is, of course, a very time consuming activity), we decided to augment the dataset. Among others, we applied rotation- and shearing operations, and modified the brightness, such that we get more training data.

The StyleGAN showed the most promising results. Starting out with a seed previously used to generate a face, we managed to train a GAN that produced something that remotely resembled art. Given that a computer is doing this, that's pretty neat!

Progressively grown StyleGAN. The adversarial network first takes on the task of creating low-resolution art, and then progressively makes sharper images (fakes).

More reading

To read a more detailed report on this project, check out the Github page:

generative-adversarial-networks

Published by:

Jeroen Overschie

You might also like...

Mar
26
RAG on GCP: production-ready GenAI on Google Cloud Platform (on Xebia.com ⧉)

RAG on GCP: production-ready GenAI on Google Cloud Platform (on Xebia.com ⧉)

GCP has an increasing set of managed services that can help you build production-ready Retrieval Augmented Generation (RAG) applications. How can you best leverage them to build RAG applications? Let’s explore together. Read along!
6 min read
Dec
28
Dataset enrichment using LLM’s ✨ (on Xebia.com ⧉)

Dataset enrichment using LLM’s ✨ (on Xebia.com ⧉)

Large Language Models (LLM's) have proven to be useful for numerous tasks. LLM's can summarise text, generate text, classify text or translate text. What LLM's can also be used for is converting unstructured data to structured data.
Apr
18
From Linear Regression to Neural Networks

From Linear Regression to Neural Networks

How are linear regression, logistic regression and neural networks related? What is overfitting and how do we fight it? In this post, we find answers to these questions in an interactive way by working with a real-world dataset on penguins.
20 min read
Oct
29
Backdoors in Neural Networks

Backdoors in Neural Networks

In this project, we demonstrated how Neural Networks can be vulnerable to a Backdoor attack.
4 min read
Mar
01
COVID-19 Dashboard

COVID-19 Dashboard

What parts of the world are susceptible to Corona outbreak? We used Big Data and Data Engineering in this project to find out.
2 min read