Tag Archives: video

biomorphic syntheses is created as a collaboration between the human image maker and machine learning image synthesis programs by training generative adversarial networks, including both styleGAN and artGAN. Renderings from these models are then used as assets in animation, reviving inanimate form to bring the appearance of life.

The project’s intention is to provoke a dialogue on how image makers can use digital programs as tools to enhance their artistic practice; as well as intending to illuminate the necessity of the creator’s input to generate the models themselves and new imagery.  At the beginning of the animation, one can see the actual visual outputs of the first steps of the artGAN learning what “surrealism” looks like. This work focuses on the idea of questioning organic design, concepts of novel being, animism, and interconnectedness of carbon systems in the forest mycorrhizal network as well as the digital network. This work creates a dialogue on carbon and pixel transfer, in both an environmental sense, data storage, and archiving.

This work is presented in a square format as both the styleGAN and artGAN output new images from transferred pixels of the inputted datasets using vector based square ratios. Deep learning systems also require a large amount of storage, subsequently requiring a great deal of energy. Presently this work has collectively required approximately 15 GB of storage. While it is difficult to measure the environmental impact of small scale deep learning models, training a single deep learning model can emit thousands of pounds of CO2 emissions.