Title: Manifold-valued Image Generation with Wasserstein Adversarial Networks

Abstract: Unsupervised image generation has recently received an increasing amount of attention thanks to the great success of generative adversarial networks (GANs), particularly Wasserstein GANs. Inspired by the paradigm of real-valued image generation, this paper makes the first attempt to formulate the problem of generating manifold-valued images, which are frequently encountered in real-world applications. For the study, we specially exploit three typical manifold-valued image generation tasks: hue-saturation-value (HSV) color image generation, chromaticity-brightness (CB) color image generation, and diffusion-tensor (DT) image generation. In order to produce such kinds of images as realistic as possible, we generalize the state-of-the-art technique of Wasserstein GANs to the manifold context with exploiting Riemannian geometry. For the proposed manifold-valued image generation problem, we recommend three benchmark datasets that are CIFAR-10 HSV/CB color images, ImageNet HSV/CB color images, UCL DT image datasets. On the three datasets, we experimentally demonstrate the proposed manifold-aware Wasserestein GAN can generate high quality manifold-valued images.