GAN Processed Still Lifes

About this series

This series is one of my tech-art projects as it uses machine learning algorithms to generate, curate and tweak the composition. I call the series "Learning from Matisse" as my initial algorithm (GAN) was trained on Matisse painted still life images. Over time, the more hand-painted still lifes I painted myself, I then updated my dataset eventually training the GAN on my own developed style. I continually return to this series as the colors and loose style are so much fun to paint.

To learn more details about the process click here.

Materials & Tools:

    • Python code
      • code to generate still life collage
      • code for classifier and cycleGAN
    • folder of painted still life images
    • paint
    • heavyweight paper

Image to the left shows digital collage (input) --> GAN processed image (output for painting) --> my painted image (appended input) ---> painterly style removed from painting