-
Manifold Adversarial Training
Manifold Adversarial Training (MAT) is a novel method to smooth the distributional manifold in the latent space. -
Toward a Visual Concept Vocabulary for GAN Latent Space
A new method for building open-ended vocabularies of primitive visual concepts represented in a GAN's latent space. -
8 Gaussians
The dataset used in the paper is a collection of data sets of varying complexity, including 8 Gaussians, MNIST, 3D Chairs, and spectrograms of zebra finch birdsong. -
Learning a probabilistic latent space of object shapes via 3D generative-adve...
Learning a probabilistic latent space of object shapes via 3D generative-adversarial modeling
