Deep learning postcards from cities hit by the Covid-19 pandemic. I wanted to capture the feelings of loss, longing and confusion by depicting destinations that hold a place in our collective imagination but that were now suddenly closed and transformed.

For this project, I went through open source archives by art museums and collected old drawings and paintings of Venice, Paris and New York. I then morphed them with new photographs of the cities by using CycleGAN. These dreamy results were mixed with images that AttnGAN created from news headlines. Texts are generated from news contents by using a RNN. Results are a mixture of dreamy, timeless depictions of the cities and the darker reality of spring 2020 that has forced us to rethink many aspects of our life.

Coded with PyTorch and TensorFlow in Python. RunwayML used for AttnGAN.