Mode matching in GANs through latent space learning and inversion
Abstract: Generative adversarial networks (GANs) have shown re-
markable success in generation of unstructured data, such
as, natural images. However, discovery and separation of
modes in the generated space, essential for several tasks
beyond naive data generation, is still a challenge. In this
paper, we address the problem of imposing desired modal
properties on the generated space using a latent distribu-
tion, engineered in accordance with the modal properties of
the true data distribution. This is achieved by training a la-
tent space inversion network in tandem with the generative
network using a divergence loss. The latent space is made
to follow a continuous multimodal distribution generated by
reparameterization of a pair of continuous and discrete ran-
dom variables. In addition, the modal priors of the latent
distribution are learned to match with the true data distri-
bution using minimal-supervision with negligible increment
in number of learnable parameters. We validate our method
on multiple tasks such as mode separation, conditional gen-
eration, and attribute discovery on multiple real world im-
age datasets and demonstrate its efficacy over other state-
of-the-art methods.
0 Replies
Loading