Autoencoded Gothic

An exercise in the use of Visual Autoencoders. Following graphic analysis, I made a linear, parametric model of a gothic window. The geometry was exported to images with pixel values changing exponentially depending on the distance from the closest line. Then I trained an Autoencoder on the generated images. The output was uploaded back into Grasshopper. The final geometry was created by extruding the pixels depending on the colour values.

2022
Software: Rhinoceros3d, Grasshopper, Python, keras, tensorflow.