Corpus ID: 211069212

Unsupervised Discovery of Interpretable Directions in the GAN Latent Space

@article{Voynov2020UnsupervisedDO,
  title={Unsupervised Discovery of Interpretable Directions in the GAN Latent Space},
  author={Andrey Voynov and Artem Babenko},
  journal={ArXiv},
  year={2020},
  volume={abs/2002.03754}
}
  • Andrey Voynov, Artem Babenko
  • Published in ArXiv 2020
  • Computer Science, Mathematics
  • The latent spaces of typical 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 can severely limit a range of directions existing approaches… CONTINUE READING

    Create an AI-powered research feed to stay up to date with new papers like this posted to ArXiv

    References

    Publications referenced by this paper.
    SHOWING 1-10 OF 34 REFERENCES

    A Style-Based Generator Architecture for Generative Adversarial Networks

    VIEW 10 EXCERPTS
    HIGHLY INFLUENTIAL

    The mnist database of handwritten digits

    VIEW 8 EXCERPTS
    HIGHLY INFLUENTIAL

    Controlling generative models with continuous factors of variations

    VIEW 8 EXCERPTS
    HIGHLY INFLUENTIAL

    GANalyze: Toward Visual Definitions of Cognitive Image Properties

    VIEW 4 EXCERPTS
    HIGHLY INFLUENTIAL

    Large Scale GAN Training for High Fidelity Natural Image Synthesis

    VIEW 5 EXCERPTS
    HIGHLY INFLUENTIAL

    On the "steerability" of generative adversarial networks

    VIEW 7 EXCERPTS
    HIGHLY INFLUENTIAL