Image-to-image translation aims at translating a particular style of an image to another. The synthesized images can be more photo-realistic and identity-preserving by decomposing the image into content and style in a disentangled manner. While existing models focus on designing specialized network architecture to separate the two components, this paper inve