Text-to-image generation on MS-COCO (zero-shot FID-30k 256x256)
6.61FIDRAPHAEL
Evaluation results
| Method | Links | ||
|---|---|---|---|
| RAPHAELType=Diff, #Params=3.0B, #Images=5000M+, GPU days=60,000 A1002023.09 | 6.61 | — | |
| ERNIE-ViLG 2.0#params=24B, denoising experts=10, reranking=true, #reranking images=4, zero-shot=true, resolution=256x2562022.10 | 6.75 | — | |
| Parti#params=20B, reranking=true, #reranking images=16, zero-shot=true, resolution=256x2562022.10 | 7.23 | — | |
| ERNIE-ViLG 2.0#params=24B, denoising experts=10, reranking=false, zero-shot=true, resolution=256x2562022.10 | 7.23 | — | |
| Imagen#params=6.6B, reranking=false, zero-shot=true, resolution=256x2562022.10 | 7.27 | — | |
| ImagenType=Diff, #Params=3.0B, #Images=860M, GPU days=7,132 A1002023.09 | 7.27 | — | |
| PIXART-aType=Diff, #Params=0.6B, #Images=25M, GPU days=753 A1002023.09 | 7.32 | — | |
| ERNIE-ViLG 2.0#params=3.5B, denoising experts=1, reranking=true, #reranking images=4, zero-shot=true, resolution=256x2562022.10 | 7.62 | — | |
| ERNIE-ViLG 2.0#params=3.5B, denoising experts=1, reranking=false, zero-shot=true, resolution=256x2562022.10 | 8.07 | — | |
| GigaGANType=GAN, #Params=0.9B, #Images=2700M, GPU days=4,783 A1002023.09 | 9.09 | — | |
| SDv1.5Type=Diff, #Params=0.9B, #Images=2000M, GPU days=6,250 A1002023.09 | 9.62 | — | |
| DALL-E 2#params=4.5B, reranking=false, zero-shot=true, resolution=256x2562022.10 | 10.39 | — | |
| DALL-E 2Type=Diff, #Params=6.5B, #Images=650M, GPU days=41,667 A1002023.09 | 10.39 | — | |
| Make-A-Scene#params=4B, reranking=false, zero-shot=true, resolution=256x2562022.10 | 11.84 | — | |
| GLIDE#params=6B, reranking=false, zero-shot=true, resolution=256x2562022.10 | 12.24 | — | |
| GLIDEType=Diff, #Params=5.0B, #Images=250M2023.09 | 12.24 | — | |
| LDM#params=1.45B, reranking=false, zero-shot=true, resolution=256x2562022.10 | 12.61 | — | |
| LDMType=Diff, #Params=1.4B, #Images=400M2023.09 | 12.64 | — | |
| DALL-E#params=12B, reranking=true, #reranking images=512, zero-shot=true, resolution=256x2562022.10 | 27.5 | — | |
| DALL-EType=Diff, #Params=12.0B, #Images=250M2023.09 | 27.5 | — | |
| DALL-E#params=12B, reranking=false, zero-shot=true, resolution=256x2562022.10 | 34.6 | — | |
| CM3Leon#params=7B, zero-shot=true, without retrieval=true2024.10 | — | 10.82 | |
| DALL-E 2#params=4.2B, zero-shot=true2024.10 | — | 10.39 | |
| Fluid#params=369M, zero-shot=true2024.10 | — | 7.23 | |
| Fluid#params=665M, zero-shot=true2024.10 | — | 6.84 | |
| Fluid#params=1.1B, zero-shot=true2024.10 | — | 6.59 | |
| Fluid#params=3.1B, zero-shot=true2024.10 | — | 6.41 | |
| Fluid#params=10.5B, zero-shot=true2024.10 | — | 6.16 | |
| Imagen#params=3B, zero-shot=true2024.10 | — | 7.27 | |
| LDM#params=1.4B, zero-shot=true2024.10 | — | 12.64 | |
| Muse#params=3B, zero-shot=true2024.10 | — | 7.88 | |
| Parti#params=20B, zero-shot=true2024.10 | — | 7.23 | |
| RAPHAEL#params=3B, zero-shot=true2024.10 | — | 6.61 | |
| Show-o#params=1.3B, zero-shot=true2024.10 | — | 9.24 | |
| Transfusion#params=7.3B, zero-shot=true2024.10 | — | 6.78 |