---

SNGAN on Anime Face Dataset

This repo provides an unofficial PyTorch Implementation of Spectral Normalization for Generative Adversarial Networks (SNGAN)¹ with specialization in Anime faces² generation.

Roadmap

• original SNGAN

  • differentiable power iteration
  • $\gamma$-reparameterization
  • ResNet architecture
  • hinge loss ³

• dataset-specific tuning

  • set $n_{dis} = 1$
  • use 1x1 conv instead of 3x3 conv for penultimate layer of G
  • replace avgpool2d down-sampling with stride-2 conv

• TTUR⁴

• mixed precision (-11% time)⁵

• cuDNN benchmark

• FID metrics + precomputed statistics

  • InceptionV3 pretrained on ImageNet is not suitable for anime data
  • data-specific feature extractor is needed

• exponential moving average

• visualizations 😄

Anime face generation

Generations from fixed noise

evolving over 50 training epochs

Interpolation

Truncation effect

from top to bottom: the variance varies from 0.3 to 3 cubically

References

---

Footnotes

1. Miyato, Takeru, et al. "Spectral Normalization for Generative Adversarial Networks." International Conference on Learning Representations. 2018. ↩

2. Chao, Brian. ‘Anime Face Dataset: a collection of high-quality anime faces’. GitHub, https://github.com/bchao1/Anime-Face-Dataset. ↩

3. Lim, Jae Hyun, and Jong Chul Ye. "Geometric gan." arXiv preprint arXiv:1705.02894 (2017). ↩

4. Heusel, Martin, et al. "Gans trained by a two time-scale update rule converge to a local nash equilibrium." Advances in neural information processing systems 30 (2017). ↩

5. "Automatic Mixed Precision For Deep Learning". NVIDIA Developer, https://developer.nvidia.com/automatic-mixed-precision. ↩