Research map / Image generation / GANs
Unsupervised Image Generation With Infinite Generative Adversarial Networks
Hui Ying, He Wang, Tianjia Shao, Yin Yang, Kun Zhou
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| use K>1 modes (MIC-GANs) instead of single GAN (K=1) (ablation) | StyleGAN2 · Hybrid (StyleGAN2) | FID: 15.47 -> 11.2 | Improved |
| use K>1 modes instead of K=1 (ablation) | DCGAN · Hybrid (DCGAN) | FID: 60.83 -> 45.07 | Improved |
| use K=15 modes instead of K=1 (ablation) | StyleGAN2 · MNIST | FID: 12.96 -> 9.92 | Improved |
| use K=15 modes instead of K=1 (ablation) | StyleGAN2 · Hybrid | FID: 15.47 -> 15.7 | No clear effect |
| vary K among 15, 20, 25 (ablation) | DCGAN · MNIST | purity: 0.9384, 0.9578, 0.93 | No clear effect |
| add MIC-GANs mode mixture to StyleGAN2-Ada | StyleGAN2-Ada · CIFAR-10 (10 classes) | FID: 5.03 -> 4.72 | Improved |
| add MIC-GANs mode mixture to StyleGAN2-Ada | StyleGAN2-Ada · CIFAR-10 (4 classes) | FID: 5.64 -> 5.31 | Improved |
| use MIC-GANs instead of Self-Conditioned GAN | Self-Conditioned GAN · CIFAR-10 (4 classes) | FID: 237.96 -> 5.31 | Improved |
| use MIC-GANs instead of DMGAN | DMGAN · MNIST | purity: 0.7612 -> 0.9489 | Improved |
| use MIC-GANs instead of DMGAN | DMGAN · Hybrid | FID: 32.31 -> 11.2 | Improved |
| use MIC-GANs instead of ClusterGAN (given ground-truth K) | ClusterGAN · MNIST | FID: 9.27 -> 12.79 | Hurt |
| use MIC-GANs instead of ClusterGAN | ClusterGAN · Hybrid | purity: 0.9457 -> 0.9567 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- DMGAN — DM-GAN: Dynamic Memory Generative Adversarial Networks for Text-To-Image Synthesis (CVPR 2019) · better on 3 of 4 shared results
- ClusterGAN — ClusterGAN: Latent Space Clustering in Generative Adversarial Networks (AAAI 2019) · better on 3 of 4 shared results
- DeliGAN — DeLiGAN : Generative Adversarial Networks for Diverse and Limited Data (CVPR 2017) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Instance-Conditioned GAN (NeurIPS 2021 · similarity 0.90)
- MSG-GAN: Multi-Scale Gradients for Generative Adversarial Networks (CVPR 2020 · similarity 0.90)
- MGAN: Training Generative Adversarial Nets with Multiple Generators (ICLR 2018 · similarity 0.89)
- Training Generative Adversarial Networks from Incomplete Observations using Factorised Discriminators (ICLR 2020 · similarity 0.89)
- Diverse Image Generation via Self-Conditioned GANs (CVPR 2020 · similarity 0.89)
- 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.
