• Computer Science, Engineering, Mathematics
  • Published in ArXiv 2019

Analyzing and Improving the Image Quality of StyleGAN

@article{Karras2019AnalyzingAI,
  title={Analyzing and Improving the Image Quality of StyleGAN},
  author={Tero Karras and Samuli Laine and Miika Aittala and Janne Hellsten and Jaakko Lehtinen and Timo Aila},
  journal={ArXiv},
  year={2019},
  volume={abs/1912.04958}
}
The style-based GAN architecture (StyleGAN) yields state-of-the-art results in data-driven unconditional generative image modeling. We expose and analyze several of its characteristic artifacts, and propose changes in both model architecture and training methods to address them. In particular, we redesign generator normalization, revisit progressive growing, and regularize the generator to encourage good conditioning in the mapping from latent vectors to images. In addition to improving image… CONTINUE READING

Citations

Publications citing this paper.
SHOWING 1-9 OF 9 CITATIONS

CNN-generated images are surprisingly easy to spot... for now

VIEW 4 EXCERPTS
CITES BACKGROUND & METHODS
HIGHLY INFLUENCED

Information Compensation for Deep Conditional Generative Networks

VIEW 11 EXCERPTS
CITES METHODS & BACKGROUND
HIGHLY INFLUENCED

Improved Consistency Regularization for GANs

VIEW 1 EXCERPT
CITES BACKGROUND

Media Forensics and DeepFakes: an overview

VIEW 1 EXCERPT
CITES BACKGROUND

References

Publications referenced by this paper.
SHOWING 1-10 OF 47 REFERENCES

A Style-Based Generator Architecture for Generative Adversarial Networks

VIEW 8 EXCERPTS

The Unreasonable Effectiveness of Deep Features as a Perceptual Metric

VIEW 5 EXCERPTS
HIGHLY INFLUENTIAL

FakeSpotter: A Simple Baseline for Spotting AI-Synthesized Fake Faces

VIEW 1 EXCERPT

Source Generator Attribution via Inversion

VIEW 1 EXCERPT