Research map / Image generation / GANs
Posterior Promoted GAN With Distribution Discriminator for Unsupervised Image Synthesis
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 |
|---|---|---|---|
| Replace the scalar discriminator with a posterior-distribution discriminator and promote generation with posterior information | P2GAN with DCGAN network settings · CIFAR10 | FID: 34.6 -> 29.7 | Helped |
| Use posterior-distribution discrimination and AdaIN promotion | P2GAN with adjusted StyleGAN structure · CelebA 32×32 | FID: 15.9 -> 15.3 | Helped |
| Use posterior-distribution discrimination and AdaIN promotion | P2GAN with adjusted StyleGAN structure · CelebA 64×64 | FID: 13.9 -> 13.4 | Helped |
| Add the latent code regularizer (ablation) | P2GAN · CIFAR10 | FID: 41.9 -> 45.6 | Helped |
| Add posterior modules (ablation) | P2GAN · CIFAR10 | FID: 38.0 -> 41.9 | Helped |
| Use the posterior latent code rather than random noise in the posterior modules (ablation) | P2GAN · CIFAR10 | FID: 35.8 -> 38.0 | Helped |
| Add the pretraining step (ablation) | P2GAN · CIFAR10 | FID: 31.1 -> 35.8 | Helped |
| Add the latent code regularizer (ablation) | P2GAN · CelebA 64×64 | FID: 12.5 -> 15.2 | Helped |
| Add posterior modules (ablation) | P2GAN · CelebA 64×64 | FID: 11.7 -> 12.5 | Helped |
| Use the posterior latent code rather than random noise in the posterior modules (ablation) | P2GAN · CelebA 64×64 | FID: 10.5 -> 11.7 | Helped |
| Add the pretraining step (ablation) | P2GAN · CelebA 64×64 | FID: 9.4 -> 10.5 | Helped |
| Use very widely separated priors N(−20.0, 1.02I) and N(20.0, 3.02I) instead of the standard priors (ablation) | P2GAN · CIFAR10 and CelebA 64×64 | FID: 31.5 -> 67.5; 9.4 -> 37.6 | Hurt |
Closest papers by meaning
- Transformation GAN for Unsupervised Image Synthesis and Representation Learning (CVPR 2020 · similarity 0.89)
- Image Processing Using Multi-Code GAN Prior (CVPR 2020 · similarity 0.89)
- Towards Faster and Stabilized GAN Training for High-fidelity Few-shot Image Synthesis (ICLR 2021 · similarity 0.88)
- Conditional Image Synthesis with Auxiliary Classifier GANs (ICML 2017 · similarity 0.88)
- MCGAN: Enhancing GAN Training with Regression-Based Generator Loss (AAAI 2025 · similarity 0.88)
- Large Scale GAN Training for High Fidelity Natural Image Synthesis (ICLR 2019 · 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.