A few experiments exploring the artistic sensibility of artificial neural networks...
These examples were generated with the Deep Dream method developed by Alexander Mordvintsev at Google. Convolutional Neural Nets, modeled on the human visual system, are useful in training a computer to distill concepts from perception through a series of filters that extract the distinct features that make objects (hence, its use in many image classification techniques). Deep Dream reverses this process undertaken by a CNN. Here, the algorithm searches for where it recognizes floral patterns in a given seed image and iteratively enhances those features. The results are these richly detailed algorithmic hallucinations.
I’m curious to find out what we can learn about human creativity from this kind of art. My master's thesis, which I completed in November 2018, explores this further. The thesis is a transdisciplinary study of topics in computer science, light art, and installation art that culminated in the design and production of two human-scale sculptural art installations, Strange Loop and in_paradiso.
Other areas I'm currently experimenting with are generative geometry, sound + image, visual style transfer, and music genre classification.