Here I have documented the creation of a runway training model starting from a GANS generated process which was trained on skyscrapers. I took images from google at random with the search result ‘file structure’. The images were not curated and just taken from the first pages of google. Then I resized them all using an online tool and fed them into runway.
Here you can see the initial steps of this with the building data beginning to be overwritten by the images of file structure.
This shows one frame from this sequence the whole way through with bellow being the whole set manipulating over the hour of training on runway. The usage of buildings within this in relevant because I have been considering data in a more materialistic way. The previous 3D print I made of data also resembled buildings so it seemed appropriate to begin this training set from this.
Here I have run the same set of images through the same training set I created before. So the initial image set and learning of the computer are based partially of skyscrapers and partially off of file structure. The way they evolve more and more to look like file structure instead is really interesting. I would like to put in other sets of file structure images in order to see the flow continuing, feeding them into the same data set which I have adapted each time. The evolving nature of the imagery is what makes it appear so entrancing. The nature of reproducing imagery over and over again is so fascinating.
oogbrain.com 