Improving Shape Deformation in Unsupervised Image-to-Image Translation
arXiv:1808.04325
Abstract
Unsupervised image-to-image translation techniques are able to map local texture between two domains, but they are typically unsuccessful when the domains require larger shape change. Inspired by semantic segmentation, we introduce a discriminator with dilated convolutions that is able to use information from across the entire image to train a more context-aware generator. This is coupled with a multi-scale perceptual loss that is better able to represent error in the underlying shape of objects. We demonstrate that this design is more capable of representing shape deformation in a challenging toy dataset, plus in complex mappings with significant dataset variation between humans, dolls, and anime faces, and between cats and dogs.
References in corpus (10)
- Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
- Rethinking Atrous Convolution for Semantic Image Segmentation
- Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks
- Conditional Random Fields as Recurrent Neural Networks
- A Learned Representation For Artistic Style
- Learning to Discover Cross-Domain Relations with Generative Adversarial Networks
- Unsupervised Cross-Domain Image Generation
- CycleGAN, a Master of Steganography
- Towards the Automatic Anime Characters Creation with Generative Adversarial Networks
- Large-Scale Deep Learning on the YFCC100M Dataset