Text-to-Image Generation on COCO 2014 (test)
3.22FIDParti-20B
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Parti-20BTraining Dataset=LAION/FIT/JFT/COCO(4.8B), Data Scale=Huge Model2024.12 | 3.22 | — | — | — | — | — | — | — | |
| Nested Diffusion ModelsTraining Dataset=COCO(83K), CFG=true, L=22024.12 | 4.72 | — | — | — | — | — | — | — | |
| CM3Leon-7BTraining Dataset=Internal Data(350M), Data Scale=Huge Model2024.12 | 4.88 | — | — | — | — | — | — | — | |
| Re-ImagenTraining Dataset=KNN-ImageText/COCO(50M), Data Scale=Huge Model2024.12 | 5.25 | — | — | — | — | — | — | — | |
| U-ViTTraining Dataset=COCO(83K), CFG=true2024.12 | 5.42 | — | — | — | — | — | — | — | |
| Nested Diffusion ModelsTraining Dataset=COCO(83K), CFG=true, L=32024.12 | 5.92 | — | — | — | — | — | — | — | |
| Nested Diffusion ModelsTraining Dataset=COCO(83K), CFG=false, L=32024.12 | 6.97 | — | — | — | — | — | — | — | |
| ImagenTraining Dataset=Internal Data/LAION (860M), Data Scale=Huge Model2024.12 | 7.27 | — | — | — | — | — | — | — | |
| Nested Diffusion ModelsTraining Dataset=COCO(83K), CFG=false, L=22024.12 | 8.15 | — | — | — | — | — | — | — | |
| FrigoTraining Dataset=COCO(83K), CFG=true2024.12 | 8.97 | — | — | — | — | — | — | — | |
| XMC-GAN2021.01 | 9.33 | 30.45 | — | 71 | 50.94 | 71.33 | — | — | |
| DALL-E 2Training Dataset=DALL-E (250M), Data Scale=Huge Model2024.12 | 10.39 | — | — | — | — | — | — | — | |
| GLIDETraining Dataset=DALL-E (250M), Data Scale=Huge Model2024.12 | 12.24 | — | — | — | — | — | — | — | |
| VQ-DiffusionTraining Dataset=COCO(83K), CFG=true2024.12 | 13.86 | — | — | — | — | — | — | — | |
| U-ViTTraining Dataset=COCO(83K), CFG=false2024.12 | 14.98 | — | — | — | — | — | — | — | |
| OP-GAN2021.01 | 24.7 | 27.88 | — | 49.8 | 35.85 | 50.47 | — | — | |
| DM-GAN2021.01 | 27.34 | 32.32 | — | — | 33.44 | 48.03 | — | — | |
| AttnGAN with VICTRwith_VICTR=true2020.10 | 29.26 | — | — | — | — | — | — | — | |
| SD-GANFID image source=shared by original paper authors2021.01 | 29.35 | 35.69 | — | 51.68 | — | — | — | — | |
| DM-GAN with VICTRwith_VICTR=true2020.10 | 32.37 | — | — | — | — | — | — | — | |
| DM-GANwith_VICTR=false2020.10 | 32.64 | — | — | — | — | — | — | — | |
| AttnGAN2021.01 | 33.1 | 23.61 | — | — | 25.88 | 39.01 | — | — | |
| AttnGANwith_VICTR=false2020.10 | 35.49 | — | — | — | — | — | — | — | |
| Obj-GAN2021.01 | 36.52 | 24.09 | — | — | 27.14 | 41.24 | — | — | |
| CP-GANFID image source=generated from open-sourced models, R-prec image source=shared by original paper authors2021.01 | 55.82 | 52.73 | — | 59.05 | 77.02 | 84.55 | — | — | |
| Full Fine-tunezero-shot=true2025.05 | — | — | — | — | — | — | 14.06 | 0.32 | |
| LoRAzero-shot=true2025.05 | — | — | — | — | — | — | 12.82 | 0.32 | |
| MaskUNetzero-shot=true2025.05 | — | — | — | — | — | — | 11.72 | 0.32 | |
| SD 1.5zero-shot=true2025.05 | — | — | — | — | — | — | 12.85 | 0.32 |