MonsterGAN
Science Fiction and horror often show what cultures are truly scared of. In the 50s and 60s, oversized creatures appeared often, like in Them!, The Deadly Mantis, and Attack of the 50 Foot Woman!, because of a fear of nuclear radiation. I began to wonder what the United State's big culture fears were over the past 50 years and if these fears would be also shown through horror and scifi. Using a DCGAN (Deep Convolutional Generative Adversarial Network), I created new movie posters based on a dataset that I created from all of IMDB's horror and science fiction movie posters. There are 2 parts to a standard DCGAN, a Generator and a Discriminator. The Generator creates fake samples that are added to the real samples. The Discriminator then tries to guess which were real and which were fake samples. The Discriminator will improve the Generator's parameters until it fools the Discriminator. After getting my new fake samples, I then used an online generator to add generative horror and scifi titles. I matched these with the images I thought that represented them best, and created a movie posters of all the fake films that you should eventually see.
2018
GitHub: