Jibbe van Schie

Learning deep learning

A attempt to think like a machine.

Learning deep learning is a project that challenges the way we think. Can we get to new or innovating outcomes by limiting our knowledge to the absolute minimum?
How would a chair design look if all we knew was 10 images chairs?

I explored ways of creating new forms and shapes from nothing but this extremely limited knowledge, and used machine learning programs to determine whether it could still be considered a chair.

When physicalizing this research I focused on the human ability to recognize a 3d object in 2d image. Creating a foldout of the chair we recognize in the image, a physical object which partially remains 2d, thus bridging both ways of thinking.

unfolded chair tappestry 1 unfolded chair tappestry 2 unfolded chair tappestry 3 unfolded chair tappestry 4 unfolded chair tappestry 5 10 images of wooden chairs stacked overlap images density map image overlap least information needed for recognition software to still recognize the image as a chair unfolded chair