Research map / Image generation / GANs
Towards Faster and Stabilized GAN Training for High-fidelity Few-shot Image Synthesis
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 SLE to the baseline generator (ablation) | Baseline GAN · Animal Face - Dog 256² | FID: 108.19 -> 94.21 | Helped |
| Add discriminator decoding/self-supervision (ablation) | Baseline GAN · Animal Face - Dog 256² | FID: 108.19 -> 56.25 | Helped |
| Add SLE and discriminator decoding together (ablation) | Baseline GAN · Animal Face - Dog 256² | FID: 108.19 -> 50.66 | Helped |
| Add discriminator decoding/self-supervision (ablation) | Baseline GAN · Art Paintings 1024² (10k images) | FID: 49.38 -> 42.53 | Helped |
| Use auto-encoding self-supervision instead of contrastive loss (ablation) | Discriminator on Art Paintings · Art Paintings (10k images) | FID: 47.14 -> 42.53 | Helped |
| Use auto-encoding self-supervision instead of aspect-ratio prediction (ablation) | Discriminator on Nature photos · Nature photos (10k images) | FID: 59.22 -> 43.65 | Helped |
| Combine auto-encoding with contrastive and aspect-ratio objectives (ablation) | Discriminator on Art Paintings · Art Paintings (10k images) | FID: 42.53 -> 44.21 | Hurt |
| Combine auto-encoding with contrastive and aspect-ratio objectives (ablation) | Discriminator on Nature photos · Nature photos (10k images) | FID: 43.65 -> 47.65 | Hurt |
| Add SLE (ablation) | Baseline generator · 1024² datasets: Flower, FFHQ, Shell, Art Paintings | FID | Helped |
| Increase training iterations (ablation) | Baseline GAN · Cat 256² | LPIPS: 2.113 -> 2.513 | Hurt |
| Increase training iterations (ablation) | Baseline GAN · Dog 256² | LPIPS: 2.073 -> 2.171 | Hurt |
| Train with StyleGAN2 finetuning rather than Ours | Few-shot GAN · Skull 1024² | FID: 130.05 -> 107.68 | Helped |
Closest papers by meaning
- Large Scale GAN Training for High Fidelity Natural Image Synthesis (ICLR 2019 · similarity 0.92)
- GAN Compression: Efficient Architectures for Interactive Conditional GANs (CVPR 2020 · similarity 0.91)
- MSG-GAN: Multi-Scale Gradients for Generative Adversarial Networks (CVPR 2020 · similarity 0.91)
- Projected GANs Converge Faster (NeurIPS 2021 · similarity 0.91)
- E$^2$GAN: Efficient Training of Efficient GANs for Image-to-Image Translation (ICML 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.