TJ Torres: Deep Style

PyData NYC 2015 I present the application of a relatively new method in unsupervised deep learning, called the Variational Autoencoder, to creating a latent feature space for image data. Using this method Ill show how we can transform images of womens tops into a latent style space and then use that space to randomly generate new synthetic images of tops. It s a harsh reality, in the world of data science, that some of the most dense and informative data is contained in unstructured sources like text o