Latent Space Conditioning on Generative Adversarial Networks
This work addresses the problem of needing annotated data for conditional GANs, offering a step towards unsupervised conditional image generation for researchers and practitioners in generative modeling.
This paper introduces a framework for unsupervised conditional Generative Adversarial Networks (GANs) that leverages representation learning to exploit the structure of a latent space. This approach allows for on-demand sample generation while maintaining the quality of supervised conditional GANs, by using unsupervised features from the latent space as conditions instead of traditional labels.
Generative adversarial networks are the state of the art approach towards learned synthetic image generation. Although early successes were mostly unsupervised, bit by bit, this trend has been superseded by approaches based on labelled data. These supervised methods allow a much finer-grained control of the output image, offering more flexibility and stability. Nevertheless, the main drawback of such models is the necessity of annotated data. In this work, we introduce an novel framework that benefits from two popular learning techniques, adversarial training and representation learning, and takes a step towards unsupervised conditional GANs. In particular, our approach exploits the structure of a latent space (learned by the representation learning) and employs it to condition the generative model. In this way, we break the traditional dependency between condition and label, substituting the latter by unsupervised features coming from the latent space. Finally, we show that this new technique is able to produce samples on demand keeping the quality of its supervised counterpart.