Research map / Image generation / GANs
Diverse Image Generation via Self-Conditioned GANs
Check 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 |
|---|---|---|---|
| derive cluster labels from discriminator features and train a conditional GAN to cover them | GAN generation on 2D-Ring · 2D-Ring | modes recovered: 7.9 ± 0.1 -> 8.0 ± 0.0 (vs. PacGAN2) | Helped |
| derive cluster labels from discriminator features and train a conditional GAN to cover them | GAN generation on 2D-Grid · 2D-Grid | reverse KL: 0.04 ± 0.01 -> 0.0063 ± 0.0007 (vs. PacGAN4) | Helped |
| derive cluster labels from discriminator features and train a conditional GAN to cover them | GAN generation on Stacked-MNIST · Stacked-MNIST | FID: 30.91 ± 1.30 -> 18.70 ± 1.28 (vs. Logo-GAN-RC) | Helped |
| derive cluster labels from discriminator features and train a conditional GAN to cover them | GAN generation on Places365 · Places365 | FID: 8.66 -> 9.76 (vs. Logo-GAN-RC) | Hurt |
| derive cluster labels from discriminator features and train a conditional GAN to cover them | GAN generation on ImageNet · ImageNet | Inception Score: 18.8583 -> 14.9620 (vs. Logo-GAN-RC) | Hurt |
| add cluster initialization using discriminator features rather than starting without it (ablation) | Self-Conditioned GAN on CIFAR-10 · CIFAR-10 | FID: 19.85 -> not reported | Helped |
| add cluster rematching after reclustering (ablation) | Self-Conditioned GAN on CIFAR-10 · CIFAR-10 | FID: 20.42 -> not reported | Helped |
| initialize clusters from random discriminator features rather than supervised class labels (ablation) | Self-Conditioned GAN on CIFAR-10 · CIFAR-10 | FID: 21.14 ± 1.59 -> not reported | Helped |
| use online clustering instead of periodic batch clustering (ablation) | Self-Conditioned GAN on Stacked-MNIST · Stacked-MNIST | modes recovered: 1000.0 ± 0.00 -> 995.8 ± 0.86 | Hurt |
| use online clustering instead of periodic batch clustering (ablation) | Self-Conditioned GAN on CIFAR-10 · CIFAR-10 | FID: not reported -> 32.45 ± 0.44 | Hurt |
Closest papers by meaning
- Diversity-Sensitive Conditional Generative Adversarial Networks (ICLR 2019 · similarity 0.90)
- Cluster-Guided Image Synthesis With Unconditional Models (CVPR 2022 · similarity 0.90)
- Diverse Rare Sample Generation with Pretrained GANs (AAAI 2025 · similarity 0.89)
- A Closer Look at Few-Shot Image Generation (CVPR 2022 · similarity 0.89)
- DeLiGAN : Generative Adversarial Networks for Diverse and Limited Data (CVPR 2017 · similarity 0.89)
- Image Generation Diversity Issues and How to Tame Them (CVPR 2025 · similarity 0.87)
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.