Installation, 2019

Digital print on plexiglass and paper, 2019

How will an AI algorithm look if it had a body? Like a robot from the movies, like a blond non threatening girl or a cute animal? If she had to choose will she choose to be female, male, animal or something in between. Something new and unexpected?

Albena Baeva teaches one deep convulsive generative adversarial network (DCGAN) to create new images by analysing tens of thousands of different bodies found online. This aims to help the algorithm find her identity by letting her create self portraits that do not necessarily resemble human fantasies of artificial intelligent creatures or man made robots.

The work is based on DCGAN by Soumith Chintala and art-DCGAN by Robbie Barrat algorithms. For the learning process it uses 3000 portraits of celebrities from CelebA Dataset, 4000 dogs from Standford Dogs Dataset, 5000 other animals from Oregon Wildlife, Animals-10, 10 Monkey Species, Google images, 14367 fashion models from Fashion Product Image Dataset, 14870 butterflies, moths, spiders and flowers from  Leeds Butterfly, Ecuador Moths, Costarica Moths, Flowers Recognition, Google Images.


photos by Bozhana Dimitrova, Albena Baeva

view from Electric dreams exhibition at gallery Plus 359

Back To Top