Research map / Image generation / GANs
DuelGAN: A Duel between Two Discriminators Stabilizes the GAN Training
Jiaheng Wei, Minghao Liu, Jiahao Luo, Andrew Zhu, James Davis, Yang Liu
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 |
|---|---|---|---|
| add Duel-D regularizer with a peer discriminator | StyleGAN-ADA · CelebA 256x256 | FID: 4.85 -> 4.32 | Improved |
| add a peer discriminator and duel game | DCGAN · CIFAR-10 | FID: 27.45 -> 21.55 | Improved |
| add a peer discriminator and duel game | DCGAN · MNIST | FID: 19.86 -> 7.87 | Improved |
| add a peer discriminator and duel game | DCGAN · VGGFace2 | FID: 49.99 -> 19.05 | Improved |
| add a peer discriminator and duel game | DCGAN · STL-10 | FID: 59.79 -> 51.37 | Improved |
| add a peer discriminator and duel game | DCGAN · CIFAR-10 | Inception score: 6.40 -> 7.45 | Improved |
| add a peer discriminator and duel game | DCGAN · STL-10 | Inception score: 5.87 -> 6.22 | Improved |
| use DuelGAN instead of Dist-GAN | GAN training · STL-10 | FID: 36.19 -> 51.37 | Hurt |
| dynamically schedule alpha and beta instead of constants (ablation) | DuelGAN · all datasets | FID/IS: approximately 10% improvement | Improved |
| increase max alpha from 0.3 to 0.5 at beta=0.5 (ablation) | DuelGAN · STL-10 | FID: 51.37 -> 58.45 | Hurt |
| increase max beta from 0.5 to 0.75 at alpha=0.3 (ablation) | DuelGAN · STL-10 | FID: 51.37 -> 59.58 | Hurt |
| lower max alpha from 0.3 to 0.1 at beta=0.5 (ablation) | DuelGAN · STL-10 | FID: 51.37 -> 58.77 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- Dist-GAN — Dist-GAN: An Improved GAN using Distance Constraints (ECCV 2018) · better on 1 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
- DO-GAN: A Double Oracle Framework for Generative Adversarial Networks (CVPR 2022 · similarity 0.91)
- Dual Discriminator Generative Adversarial Nets (NeurIPS 2017 · similarity 0.91)
- GLeaD: Improving GANs With a Generator-Leading Task (CVPR 2023 · similarity 0.90)
- AlphaGAN: Fully Differentiable Architecture Search for Generative Adversarial Networks (TPAMI 2021 · similarity 0.89)
- Projected GANs Converge Faster (NeurIPS 2021 · similarity 0.89)
- MCGAN: Enhancing GAN Training with Regression-Based Generator Loss (AAAI 2025 · 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.
