Use a GAN to create new human faces.
The idea of this project was to use a Generative Adversial Network (GAN) to create new content.
This was my fifth and last project of the Deep Learning Nanodegree. With this project we were asked to create a generative adversarial networks (GAN). The first neural network had to create faces using a random vector and the second one guess which faces were fake and which ones were real.
The faces generated looks enough realistic. The dataset used to train the network is big (it have around 200k images) so in order to reduce training times I cropped the images to 28 x 28 pixels. In the future I would like to train it without the cropping to see how the network performs.