Research map / Image generation / GANs
Deceive D: Adaptive Pseudo Augmentation for GAN Training with Limited Data
Liming Jiang, Bo Dai, Wayne Wu, Chen Change Loy
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 |
|---|---|---|---|
| Apply APA to StyleGAN2 | StyleGAN2 · FFHQ-5k | FID: 37.830 -> 13.249 | Improved |
| Apply APA to StyleGAN2 | StyleGAN2 · AFHQ-Cat-5k | FID: 7.737 -> 4.876 | Improved |
| Apply APA to StyleGAN2 | StyleGAN2 · Anime-5k | FID: 23.778 -> 13.089 | Improved |
| Apply APA to StyleGAN2 | StyleGAN2 · CUB-12k | FID: 23.437 -> 12.889 | Improved |
| Apply APA to StyleGAN2 | StyleGAN2 · FFHQ-70k | FID: 3.862 -> 3.678 | Improved |
| Combine APA with ADA | StyleGAN2 · FFHQ-5k | FID: 11.409 -> 8.379 | Improved |
| Use lambda_f instead of lambda_r (ablation) | APA on StyleGAN2 · FFHQ-5k | FID: 13.249 -> 13.470 | Hurt |
| Use lambda_rf instead of lambda_r (ablation) | APA on StyleGAN2 · FFHQ-5k | FID: 13.249 -> 12.679 | Improved |
| Fix deception probability at p=0.5 instead of adapting it (ablation) | APA on StyleGAN2 · FFHQ-5k | FID: 13.249 -> 14.632 | Hurt |
| Use two-sided label flipping instead of the main deception strategy (ablation) | APA on StyleGAN2 · FFHQ-5k | FID: 13.249 -> 15.440 | Hurt |
| Set threshold t to 0.4 instead of 0.6 (ablation) | APA on StyleGAN2 · FFHQ-5k | FID: 13.249 -> 13.687 | Hurt |
| Set threshold t to 0.8 instead of 0.6 (ablation) | APA on StyleGAN2 · FFHQ-5k | FID: 13.249 -> 13.984 | Hurt |
Closest papers by meaning
- Training Generative Adversarial Networks with Limited Data (NeurIPS 2020 · similarity 0.89)
- Augmentation-Aware Self-Supervision for Data-Efficient GAN Training (NeurIPS 2023 · similarity 0.89)
- Differentiable Augmentation for Data-Efficient GAN Training (NeurIPS 2020 · similarity 0.88)
- NICE: NoIse-modulated Consistency rEgularization for Data-Efficient GANs (NeurIPS 2023 · similarity 0.88)
- Improving the Training of Data-Efficient GANs via Quality Aware Dynamic Discriminator Rejection Sampling (CVPR 2025 · similarity 0.87)
- MCGAN: Enhancing GAN Training with Regression-Based Generator Loss (AAAI 2025 · similarity 0.87)
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.
