A Unified Feature Disentangler for Multi-Domain Image Translation and\n Manipulation

We present a novel and unified deep learning framework which is capable of\nlearning domain-invariant representation from data across multiple domains.\nRealized by adversarial training with additional ability to exploit\ndomain-specific information, the proposed network is able to perform continuous\ncross-domain image translation and manipulation, and produces desirable output\nimages accordingly. In addition, the resulting feature representation exhibits\nsuperior performance of unsupervised domain adaptation, which also verifies the\neffectiveness of the proposed model in learning disentangled features for\ndescribing cross-domain data.\n