Research map / Label-efficient and robust learning / Self-supervised and contrastive
Self-Supervised GANs with Label Augmentation
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 |
|---|---|---|---|
| Replace separate GAN and transformation-classification objectives with a unified label-augmented discriminator | BigGAN · CIFAR-10 | FID: 10.83 -> 5.87 | Helped |
| Replace separate GAN and transformation-classification objectives with a unified label-augmented discriminator | BigGAN · STL-10 | FID: 20.15 -> 14.58 | Helped |
| Replace separate GAN and transformation-classification objectives with a unified label-augmented discriminator | BigGAN · Tiny-ImageNet | FID: 31.01 -> 23.69 | Helped |
| Add label-augmented self-supervision to BigGAN + DiffAugment | BigGAN + DiffAugment · CIFAR-10 100% data | FID: 8.70 -> 8.05 | Helped |
| Add label-augmented self-supervision to BigGAN + DiffAugment | BigGAN + DiffAugment · CIFAR-10 20% data | FID: 14.04 -> 11.16 | Helped |
| Add label-augmented self-supervision to BigGAN + DiffAugment | BigGAN + DiffAugment · CIFAR-100 10% data | FID: 33.70 -> 23.17 | Helped |
| Retain the original discriminator alongside the label-augmented discriminator (ablation) | SSGAN-LA · CIFAR-10 | FID: 5.87 -> 6.64 | Hurt |
| Retain the original discriminator alongside the label-augmented discriminator (ablation) | SSGAN-LA · STL-10 | FID: 14.58 -> 15.91 | Hurt |
| Increase SSGAN generator self-supervision weight from 0.0 to the suggested value (ablation) | SSGAN · CIFAR-10 | FID: 8.07 -> 7.52 | Helped |
| Increase SSGAN generator self-supervision weight from the suggested value to 1.0 (ablation) | SSGAN · STL-10 | FID: 16.84 -> 18.88 | Hurt |
| Increase SSGAN-MS generator self-supervision weight from 0.0 to the suggested value (ablation) | SSGAN-MS · Tiny-ImageNet | FID: 99.74 -> 25.76 | Helped |
| Increase SSGAN-MS generator self-supervision weight from the suggested value to 1.0 (ablation) | SSGAN-MS · STL-10 | FID: 16.46 -> 19.26 | Hurt |
Closest papers by meaning
- Augmentation-Aware Self-Supervision for Data-Efficient GAN Training (NeurIPS 2023 · similarity 0.93)
- A Unified Generative Adversarial Network Training via Self-Labeling and Self-Attention (ICML 2021 · similarity 0.93)
- Self-supervised GAN: Analysis and Improvement with Multi-class Minimax Game (NeurIPS 2019 · similarity 0.92)
- Self-Supervised GANs via Auxiliary Rotation Loss (CVPR 2019 · similarity 0.89)
- Fusing Conditional Submodular GAN and Programmatic Weak Supervision (AAAI 2024 · similarity 0.86)
- MCGAN: Enhancing GAN Training with Regression-Based Generator Loss (AAAI 2025 · similarity 0.86)
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.