I am an artist and researcher who likes to work with large data sources that I create myself to construct narratives. I try to use new technologies and am particularly interested in machine learning.
Fall of the House of Usher
Fall of the House of Usher is a 12 minute animation that I made, where each still is generated using a neural net (pix2pix, a deep-learning image-to-image translation system) that I trained on my own drawings. To create this I made my own training set of 200 drawings of stills from the 1929 film version of Fall of House of Usher as my base.This lead to the film. As the training set came from the first 4 minutes, as the film progresses, the information starts to break down. There are no training images so the programme is having to construct every frame from what it already knows. The net I have made can now draw in my style, but because its training set was only made up with early stills of the film, it has a limited vocabulary.
Drawing with Sound
This is a durational performance project has turned the act drawing into a musical instrument, made with composer Ben Heim from the Royal College of Music. I have trained a neural net to recognise shapes that I commonly make from a data set made up of my life-drawing sketches. When I perform I wear a pair of glasses with a webcam attached: when a shape is recognised it triggers the sound of a soprano's voice to played in real time. The performance occurs over a period of time, with the track drawn in charcoal on a white wall and then erased, echoing the repetitions that a machine learning programme goes through in different epochs when training and learning. Because each track is redrawn in the same space, traces build up over time, which causes disorder, and, ultimately entropy - a track drawn at the start will be clearer and more precise than one made towards the end.