If my painiting robot, Zanelle, were self-aware, I think that she would like painting other robots.
Currently, I am (of course) picking the subject matter that she paints, and I always get a kick out of it when I have her paint other robots. Maybe this joke is already old, or will get old soon, but I like the resulting artifact regardless.
There is also a lot more going on then simple painting. I have written several neural net algorithms that make aesthetic decisions on behalf of Zanelle. Below are two examples of R2-D2. Underneath each post I briefly discuss what I provided as input and also what Zanelle’s algorithms “interpreted” from my input. As you read this, think of Zanelle as a complex generative art system. I give her a seed, in most cases an image, and based on that seed she follows several rules that I have created to produce a variety of art.
In the first painting, all I provided was a picture of R2-D2 scooting along. I flattened the background of the picture and made it monotone. Zanelle’s algorithms analyzed the image, recognized the figure, and separated the foreground from the background using K-Means clustering. Zanelle then selected a crosshatching algorithm and the color orange to depict the background. It also selected the color blue to paint R2-D2. Finally, it requested that I provide line art of the figure. In Zanelle’s painting interface, I traced R2-D2. I hope to automate line drawing shortly, but for now it is manual. Anyways, once this was complete, Zanelle painted nine renditions with an artists brush and acrylic. She painted the images on wood. I then mounted the nine wood panels on a larger panel and the piece was complete. This work, and two others, was recently accepted into the FLIK Interactive Exhibition in Washington D.C. (www.artoutlet.org).
In this piece, which we just finished, Zanelle had a lot more autonomy. The only thing I provided was a frontal image of R2-D2. Zanelle made all stylistic decisions from there. First off, her algorithms used K-Means clustering to reduce the colors to 4. Zanelle then decided, based on my neural net algorithms, to change the colors so that the R2 Unit was painted in three colors ranging from white to dark blue. Red was selected as the background. 5 panels were painted in the scheme before Zanelle started scrambling the colors. The scrambling was experimental and continued until 16 panels were painted. Once this occurred, I photographed the 16 panels and provided it as feedback into one last Zanelle algorithm. This last algorithm selected what it considered the ultimate arrangement. I mounted the arrangement as instructed.
This has been two examples of how my robot and I make art together.
Pindar