Research map / Image generation / GANs
Dist-GAN: An Improved GAN using Distance Constraints
Ngoc-Trung Tran, Tuan-Anh Bui, Ngai-Man Cheung
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 latent-data distance constraint L_W (ablation) | Dist-GAN · CIFAR-10 ablation (CNN) | FID: 36.09 -> 28.23 | Improved |
| add discriminator-score distance generator objective L_G (ablation) | Dist-GAN · CIFAR-10 ablation (CNN) | FID: >200 -> 28.23 | Improved |
| add reconstructed samples as real term L_C (ablation) | Dist-GAN · CIFAR-10 ablation (CNN) | FID: 29.80 -> 28.23 | Improved |
| add gradient penalty L_P (ablation) | Dist-GAN · CIFAR-10 ablation (CNN) | FID: 28.50 -> 28.23 | Improved |
| add both L_C and L_P (ablation) | Dist-GAN · CIFAR-10 ablation (CNN) | FID: 30.37 -> 28.23 | Improved |
| treat reconstructed samples as real rather than fake (ablation) | Dist-GAN1 · Synthetic 25 Gaussians | registered modes: 17.67 -> 23.62 | Improved |
| treat reconstructed samples as real rather than none (ablation) | Dist-GAN1 · Synthetic 25 Gaussians | registered modes: 21.96 -> 23.62 | Improved |
| add gradient penalty to Dist-GAN1 (real) (ablation) | Dist-GAN1 · Synthetic 25 Gaussians | TV (Differential): 0.80 -> 0.18 | Improved |
| add latent-data distance constraint to Dist-GAN2 (ablation) | Dist-GAN2 · Synthetic 25 Gaussians | TV (Differential): 0.18 -> 0.12 | Improved |
| add discriminator-score generator objective to baseline GAN (ablation) | GAN · Synthetic 25 Gaussians | registered modes: 21.50 -> 23.75 | Improved |
| add gradient penalty to GAN1 (ablation) | GAN1 · Synthetic 25 Gaussians | TV (Differential): 0.86 -> 0.33 | Improved |
| replace WGAN-GP with Dist-GAN | WGAN-GP · MNIST-1K (K/4 discriminator) | covered modes: 640.1 -> 859.5 | Improved |
Later papers that compared with this method
3 later accepted papers report Dist-GAN as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- DuelGAN: A Duel between Two Discriminators Stabilizes the GAN Training (ECCV 2022 · better on 1 of 2 shared results)
- Dynamically Grown Generative Adversarial Networks (AAAI 2021 · better on 1 of 1 shared result)
- Improving GAN with Neighbors Embedding and Gradient Matching (AAAI 2019 · 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
- Improving GAN with Neighbors Embedding and Gradient Matching (AAAI 2019 · similarity 0.91)
- Constrained Generative Adversarial Networks for Interactive Image Generation (CVPR 2019 · similarity 0.90)
- Training Generative Adversarial Networks in One Stage (CVPR 2021 · similarity 0.90)
- Mixed batches and symmetric discriminators for GAN training (ICML 2018 · similarity 0.90)
- Dual Discriminator Generative Adversarial Nets (NeurIPS 2017 · similarity 0.90)
- 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.
