Deep Empty - Wide Open
Collaboration with Valerie Messini.
The installation is based on an idea by Valerie Messini which is outlined here. In short, we use Generative Adversarial Networks (i.e., an architecture called CycleGAN) to learn a mapping from pictures of humans to pictures of sea horizons in a semi-supervised way. In the live installation, live videos of visitors are then transformed into sea horizon images.
The resulting images/videos are somewhat reminiscent of impressionist paintings, and the installation invites visitors to interactively explore the generative capabilities of the algorithm with facial expressions and gestures.
Exhibited at Artemis Gallery Lisbon in the exhibition Beyond Human.