Research map / Image generation / GANs
Improving GANs with A Dynamic Discriminator
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 decreasing-capacity schedule (full to half width) to the discriminator (ablation) | StyleGAN2 · FFHQ-0.1K | FID: 179.21 -> 50.37 | Helped |
| add decreasing-capacity schedule to the discriminator (ablation) | StyleGAN2 · FFHQ-2K | FID: 78.82 -> 23.47 | Helped |
| add increasing-capacity schedule (half to full width) (ablation) | StyleGAN2 · FFHQ-140K | FID: 3.75 -> 3.53 | Helped |
| use increasing capacity on limited data (wrong direction) (ablation) | StyleGAN2 · FFHQ-0.1K | FID: 179.21 -> 181.03 | No clear effect |
| use decreasing capacity on sufficient data (wrong direction) (ablation) | StyleGAN2 · FFHQ-140K | FID: 3.75 -> 3.74 | No clear effect |
| fixed half-width discriminator instead of full (ablation) | StyleGAN2 · FFHQ-0.1K / 2K / 140K | FID: 179.21 -> 137.31 (0.1K); 3.75 -> 4.73 (140K) | Mixed |
| add DynamicD to StyleGAN2 on AFHQ and LSUN | StyleGAN2 · AFHQ cat/dog/wild, LSUN church/bridge/bedroom | FID: 5.65 -> 4.01 (Bedroom) | Helped |
| add DynamicD to StyleNeRF | StyleNeRF · FFHQ-2K, FFHQ-140K, Carla-2K, Carla-10K | FID: 73.50 -> 23.29 (FFHQ-2K) | Helped |
| add DynamicD to ADA | ADA on StyleGAN2 · FFHQ-0.1K | FID: 82.17 -> 62.30 | Helped |
| add DynamicD to zCR | zCR on StyleGAN2 · FFHQ-0.1K | FID: 179.66 -> 66.01 | Helped |
| add DynamicD during fine-tuning | StyleGAN2 fine-tuning · MetFaces | FID: 22.93 -> 20.52 | Helped |
Closest papers by meaning
- Dynamically Masked Discriminator for GANs (NeurIPS 2023 · similarity 0.92)
- Improving the Training of Data-Efficient GANs via Quality Aware Dynamic Discriminator Rejection Sampling (CVPR 2025 · similarity 0.91)
- Dynamically Grown Generative Adversarial Networks (AAAI 2021 · similarity 0.91)
- Training Generative Adversarial Networks with Limited Data (NeurIPS 2020 · similarity 0.90)
- Improving Generalization and Stability of Generative Adversarial Networks (ICLR 2019 · similarity 0.90)
- Discriminative Forests Improve Generative Diversity for Generative Adversarial Networks (AAAI 2024 · 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.