Research map / Image generation / GANs
GenCo: Generative Co-training for Generative Adversarial Networks with Limited Data
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 GenCo co-training | BigGAN · CIFAR-10 10% data | FID: 48.08 $\pm$0.10 -> 28.08 $\pm$0.11 | Helped |
| Add GenCo co-training | StyleGAN2 · 100-shot Obama | FID: 80.16 $\pm$0.22 -> 36.28 $\pm$0.11 | Helped |
| Add WeCo (weight-diversified discriminators) (ablation) | BigGAN · CIFAR-10 10% data | FID: 48.08 $\pm$0.10 -> 34.05 $\pm$0.15 | Helped |
| Add DaCo (different frequency-component input views) (ablation) | BigGAN · CIFAR-10 10% data | FID: 48.08 $\pm$0.10 -> 30.33 $\pm$0.13 | Helped |
| Combine WeCo and DaCo (ablation) | BigGAN · CIFAR-10 10% data | FID: 34.05 $\pm$0.15 -> 28.08 $\pm$0.11 | Helped |
| Combine WeCo and DaCo (ablation) | StyleGAN2 · 100-shot Obama | FID: 41.96 $\pm$0.19 -> 36.28 $\pm$0.11 | Helped |
| Use random frequency rejection as augmentation without co-training (ablation) | BigGAN · CIFAR-10 10% data | FID: 48.08 $\pm$0.10 -> 40.36 $\pm$0.11 | Helped |
| Use DaCo co-training rather than frequency rejection as augmentation alone (ablation) | BigGAN · CIFAR-10 10% data | FID: 40.36 $\pm$0.11 -> 30.33 $\pm$0.13 | Helped |
| Set rejected frequency-component percentage to 0.2 rather than 0.1 (ablation) | DaCo · CIFAR-10 10% data | FID: 33.02 $\pm$0.09 -> 28.08 $\pm$0.11 | Helped |
| Vary the rejected frequency-component percentage across 0.1–0.5 (ablation) | DaCo · CIFAR-10 10% data | IS: 7.76 $\pm$0.20; 7.94 $\pm$026; 7.83 $\pm$0.17; 7.88 $\pm$0.18 | No clear effect |
| Add GenCo to DA augmentation | BigGAN · CIFAR-10 10% data | FID: 23.34 $\pm$0.09 -> 18.10 $\pm$0.06 | Helped |
| Add GenCo to LeCam-GAN | BigGAN · CIFAR-10 10% data | FID: 35.23 $\pm$0.14 -> 25.89 $\pm$0.07 | Helped |
Closest papers by meaning
- Training Generative Adversarial Networks with Limited Data (NeurIPS 2020 · similarity 0.89)
- Towards Principled Methods for Training Generative Adversarial Networks (ICLR 2017 · similarity 0.89)
- MCL-GAN: Generative Adversarial Networks with Multiple Specialized Discriminators (NeurIPS 2022 · similarity 0.89)
- MGAN: Training Generative Adversarial Nets with Multiple Generators (ICLR 2018 · similarity 0.89)
- Training Generative Adversarial Networks in One Stage (CVPR 2021 · similarity 0.88)
- MCGAN: Enhancing GAN Training with Regression-Based Generator Loss (AAAI 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.