Research map / Image generation / GANs
GLeaD: Improving GANs With a Generator-Leading Task
Read the paperCheck 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 |
|---|---|---|---|
| Add generator-leading reconstruction training to StyleGAN2's discriminator | StyleGAN2 · LSUN Bedroom | FID: 4.30 -> 2.55 | Helped |
| Add generator-leading reconstruction training to StyleGAN2's discriminator | StyleGAN2 · LSUN Church | FID: 4.04 -> 2.82 | Helped |
| Add generator-leading reconstruction training to StyleGAN2's discriminator | StyleGAN2 · LSUN Bedroom | Recall: 0.25 -> 0.39 | Helped |
| Combine GLeaD with GGDR | StyleGAN2 · FFHQ, LSUN Bedroom, LSUN Church | FID: 3.25 -> 2.90; 3.71 -> 2.72; 2.81 -> 2.15 | Helped |
| Disable both real- and fake-image reconstruction tasks (ablation) | GLeaD · LSUN Church | FID: 2.82 -> 4.04 | Hurt |
| Set real-image reconstruction weight to 100 with fake-image reconstruction disabled (ablation) | GLeaD · LSUN Church | FID: 3.10 -> 331 | Hurt |
| Add real-image reconstruction only with λ1=10 (ablation) | GLeaD · LSUN Church | FID: 4.04 -> 3.10 | Helped |
| Add fake-image reconstruction only with λ2=10 (ablation) | GLeaD · LSUN Church | FID: 4.04 -> 3.32 | Helped |
| Add fake-image reconstruction with λ2=3 alongside real-image reconstruction at λ1=10 (ablation) | GLeaD · LSUN Church | FID: 3.10 -> 2.82 | Helped |
| Use 1×1 features, removing spatial information and retaining latent codes only (ablation) | GLeaD · LSUN Church | FID: 4.04 -> 4.68 | Hurt |
| Use 32×32 rather than 16×16 spatial features (ablation) | GLeaD · LSUN Church | FID: 3.01 -> 2.82 | Helped |
| Use 64×64 rather than 32×32 spatial features (ablation) | GLeaD · LSUN Church | FID: 2.82 -> 2.88 | Hurt |
Closest papers by meaning
- Are GANs Created Equal? A Large-Scale Study (NeurIPS 2018 · similarity 0.91)
- Training Generative Adversarial Networks in One Stage (CVPR 2021 · similarity 0.91)
- MGAN: Training Generative Adversarial Nets with Multiple Generators (ICLR 2018 · similarity 0.90)
- Training Generative Adversarial Networks with Limited Data (NeurIPS 2020 · similarity 0.90)
- RG-GAN: Dynamic Regenerative Pruning for Data-Efficient Generative Adversarial Networks (AAAI 2024 · 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.