Unsupervised Discovery of Interpretable Directions in the GAN Latent Space
This addresses the need for more controllable GAN generation without reliance on labels or pretrained models, offering a novel approach with practical applications in image editing and saliency detection.
The paper tackles the problem of discovering interpretable directions in GAN latent spaces without supervision, introducing an unsupervised method that identifies directions for semantic manipulations like background removal and achieves competitive performance in weakly-supervised saliency detection.
The latent spaces of GAN models often have semantically meaningful directions. Moving in these directions corresponds to human-interpretable image transformations, such as zooming or recoloring, enabling a more controllable generation process. However, the discovery of such directions is currently performed in a supervised manner, requiring human labels, pretrained models, or some form of self-supervision. These requirements severely restrict a range of directions existing approaches can discover. In this paper, we introduce an unsupervised method to identify interpretable directions in the latent space of a pretrained GAN model. By a simple model-agnostic procedure, we find directions corresponding to sensible semantic manipulations without any form of (self-)supervision. Furthermore, we reveal several non-trivial findings, which would be difficult to obtain by existing methods, e.g., a direction corresponding to background removal. As an immediate practical benefit of our work, we show how to exploit this finding to achieve competitive performance for weakly-supervised saliency detection.