Text-to-Image Generation on SketchDUO (test)
143.68FIDStable Diffusion v1.5 + VAE fine-tuning
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Stable Diffusion v1.5 + VAE fine-tuningBase Model Version=v1.5, Fine-tuning Protocol=U-Net + VAE fine-tuning2025.10 | 143.68 | 0.37 | 35.48 | 0.88 | 0.68 | |
| Stable Diffusion v2.1 + U-Net fine-tuningBase Model Version=v2.1, Fine-tuning Protocol=U-Net fine-tuning2025.10 | 144.46 | 0.41 | 34.79 | 0.88 | 0.67 | |
| Stable Diffusion v1.5 + U-Net fine-tuningBase Model Version=v1.5, Fine-tuning Protocol=U-Net fine-tuning2025.10 | 161.94 | 0.4 | 36.05 | 0.89 | 0.68 | |
| Stable Diffusion v2.1 + VAE fine-tuningBase Model Version=v2.1, Fine-tuning Protocol=U-Net + VAE fine-tuning2025.10 | 172.35 | 0.5 | 34.11 | 0.88 | 0.65 | |
| Stable Diffusion v1.5Base Model Version=v1.5, Fine-tuning Protocol=None2025.10 | 207.59 | 0.69 | 34 | 0.89 | 0.59 | |
| Stable Diffusion v2.1Base Model Version=v2.1, Fine-tuning Protocol=None2025.10 | 230.78 | 0.72 | 31.13 | 0.88 | 0.53 |