Learning Privacy Preserving Encodings through Adversarial Training