Twin-GAN -- Unpaired Cross-Domain Image Translation with Weight-Sharing GANs
This addresses the problem of generating realistic images across domains without paired data for applications like face editing, though it appears incremental as it builds on existing GAN-based translation methods.
The paper tackles unpaired cross-domain image translation by proposing Twin-GAN, a framework using weight-sharing GANs with multiple losses, and demonstrates its capability on face image translation without supervised mappings.
We present a framework for translating unlabeled images from one domain into analog images in another domain. We employ a progressively growing skip-connected encoder-generator structure and train it with a GAN loss for realistic output, a cycle consistency loss for maintaining same-domain translation identity, and a semantic consistency loss that encourages the network to keep the input semantic features in the output. We apply our framework on the task of translating face images, and show that it is capable of learning semantic mappings for face images with no supervised one-to-one image mapping.