Research map / Image generation / GANs
Progressive Augmentation of GANs
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Add progressive input-space augmentation | SN DCGAN · CIFAR10 | FID: 26.0 -> 22.2 | Helped |
| Add progressive feature-space augmentation | SN DCGAN · CELEBA-HQ | FID: 24.3 -> 18.8 | Helped |
| Use progressive augmentation rather than single-level augmentation (ablation) | SN DCGAN · GAN image generation | FID | Helped |
| Add feature-space augmentation at N/2 rather than N/8 (ablation) | SN DCGAN · CIFAR10 | FID: 22.6 -> 22.8 | Hurt |
| Add feature-space augmentation at N/2 rather than N/8 (ablation) | SN DCGAN · CELEBA-HQ | FID: 18.8 -> 19.6 | Hurt |
| Combine input-space PA with dropout | SN DCGAN · CIFAR10 | FID: 22.2 -> 21.9 | Helped |
| Combine PA with self-supervised rotation training | SA GAN (sBN) · CIFAR10 | FID: 16.1 -> 14.7 | Helped |
| Add gradient penalty to PA | SN DCGAN · CIFAR10 | FID: 22.2 -> 21.8 | Hurt |
| Add one-sided label smoothing to PA | SN DCGAN · CIFAR10 | FID: 22.2 -> 23.1 | Hurt |
| Replace entry-wise dropout with spatial dropout (ablation) | SN DCGAN · CIFAR10 | FID: 22.1 -> 23.4 | Hurt |
| Use the progression scheduler for learning-rate adaptation instead of PA (ablation) | SN DCGAN · CIFAR10 | FID: 22.2 vs. 24.0 | Hurt |
Closest papers by meaning
- A Large-Scale Study on Regularization and Normalization in GANs (ICML 2019 · similarity 0.91)
- Progressive Growing of GANs for Improved Quality, Stability, and Variation (ICLR 2018 · similarity 0.91)
- Stabilizing Training of Generative Adversarial Networks through Regularization (NeurIPS 2017 · similarity 0.90)
- On gradient regularizers for MMD GANs (NeurIPS 2018 · similarity 0.90)
- Projected GANs Converge Faster (NeurIPS 2021 · similarity 0.90)
- Differentiable Augmentation for Data-Efficient GAN Training (NeurIPS 2020 · similarity 0.89)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.