（原）Show, Attend and Translate: Unsupervised Image Translation with Self-Regularization and Attention