Research map / Image generation / GANs
Scaling Up GANs for Text-to-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 |
|---|---|---|---|
| increase the text-conditioned StyleGAN2 baseline width 5.7× (ablation) | StyleGAN2 text-to-image baseline · 64px text-to-image | FID-10k: 29.91 -> 34.07 | Hurt |
| tune the StyleGAN2 configuration (ablation) | StyleGAN2 text-to-image baseline · 64px text-to-image | FID-10k: 29.91 -> 28.11 | Helped |
| add attention (ablation) | Tuned StyleGAN2 · 64px text-to-image | FID-10k: 28.11 -> 23.87 | Helped |
| add matching-aware loss to the discriminator only (ablation) | Attention model · 64px text-to-image | FID-10k: 23.87 -> 27.29 | Hurt |
| apply matching-aware loss to both generator and discriminator (ablation) | Attention model · 64px text-to-image | FID-10k: 23.87 -> 21.66 | Helped |
| add adaptive convolution (ablation) | Matching-aware G and D model · 64px text-to-image | FID-10k: 21.66 -> 19.97 | Helped |
| increase generator depth (ablation) | Adaptive convolution model · 64px text-to-image | FID-10k: 19.97 -> 19.18 | Helped |
| add CLIP loss (ablation) | Deeper model · 64px text-to-image | FID-10k: 19.18 -> 14.88 | Helped |
| add multi-scale training (ablation) | CLIP-loss model · 64px text-to-image | FID-10k: 14.88 -> 14.92 | Hurt |
| add vision-aided GAN loss (ablation) | Multi-scale training model · 64px text-to-image | FID-10k: 14.92 -> 13.67 | Helped |
| scale up the final formulation and training (ablation) | Vision-aided GAN model · 64px text-to-image | FID-10k: 13.67 -> 9.18 | Helped |
| add multi-scale training (ablation) | CLIP-loss model · 64px text-to-image | CLIP Score: 0.280 -> 0.300 | Helped |
Closest papers by meaning
- StyleGAN-T: Unlocking the Power of GANs for Fast Large-Scale Text-to-Image Synthesis (ICML 2023 · similarity 0.94)
- Large Scale GAN Training for High Fidelity Natural Image Synthesis (ICLR 2019 · similarity 0.90)
- E$^2$GAN: Efficient Training of Efficient GANs for Image-to-Image Translation (ICML 2024 · similarity 0.90)
- The GAN is dead; long live the GAN! A Modern GAN Baseline (NeurIPS 2024 · similarity 0.88)
- Adversarial Text to Continuous Image Generation (CVPR 2024 · similarity 0.88)
- Exploring Sparse MoE in GANs for Text-conditioned Image Synthesis (CVPR 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.