Research map / Image generation / GANs
Self-Diagnosing GAN: Diagnosing Underrepresented Samples in Generative Adversarial Networks
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 score-based weighted sampling of real samples (Dia-GAN) on top of vanilla GAN | SNGAN · CIFAR-10 | FID: 26.90 -> 19.66 | Helped |
| add Dia-GAN weighted sampling beyond DRS post-processing alone (ablation) | SNGAN · CIFAR-10 | FID: 24.54 -> 19.66 | Helped |
| Dia-GAN weighted sampling | SSGAN · CIFAR-10 | FID: 22.01 -> 16.31 | Helped |
| Dia-GAN weighted sampling | SNGAN · CelebA | recall: 0.44 -> 0.48 | Helped |
| Dia-GAN weighted sampling | SNGAN · CelebA | precision: 0.68 -> 0.64 | Hurt |
| reweight fake samples (GOLD) instead of real samples | SNGAN · CIFAR-10 | FID: 26.90 -> 28.86 | Hurt |
| Dia-GAN weighted sampling | StyleGAN2 · FFHQ 256x256 | FID: 14.07 -> 11.89 | Helped |
| apply Dia sampling using hinge-loss discriminator output statistics | HingeGAN · CIFAR-10 | FID: 21.99 -> 18.74 | Helped |
| apply Dia sampling using hinge-loss discriminator output statistics | HingeGAN · CelebA | FID: 6.66 -> 5.98 | Helped |
| Dia-GAN emphasis of underrepresented samples at 99% majority rate | DCGAN · Colored MNIST minor RE | reconstruction error: 0.838 -> 0.224 | Helped |
| Dia-GAN emphasis of underrepresented samples at 90% majority rate | DCGAN · MNIST-FMNIST minor RE | reconstruction error: 0.215 -> 0.206 | Helped |
| Dia-GAN emphasis on minor CelebA attributes | SNGAN · CelebA Bald | partial recall: 0.353 -> 0.393 | Helped |
Closest papers by meaning
- AdaGAN: Boosting Generative Models (NeurIPS 2017 · similarity 0.89)
- Instance Selection for GANs (NeurIPS 2020 · similarity 0.89)
- Consistency-GAN: Training GANs with Consistency Model (AAAI 2024 · similarity 0.88)
- MCGAN: Enhancing GAN Training with Regression-Based Generator Loss (AAAI 2025 · similarity 0.88)
- Diverse Rare Sample Generation with Pretrained GANs (AAAI 2025 · similarity 0.88)
- Improving the Training of Data-Efficient GANs via Quality Aware Dynamic Discriminator Rejection Sampling (CVPR 2025 · similarity 0.88)
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.