Text-to-Image Generation on MS-COCO 30K (test)
4.88FIDCM3Leon-7B
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CM3Leon-7BRetrieval in Training=true, Responsible=true, Number of Retrieved Documents=2, Dataset Size=340M, Model Size=7B, Zero-shot=true, Samples per query=8, Selection=CLIP-based re-ranking2023.09 | 4.88 | — | |
| RE-IMAGENRetrieval in Training=true, Responsible=false, Number of Retrieved Documents=2, Dataset Size=150M, Model Size=3.6B, Zero-shot=true2023.09 | 5.25 | — | |
| CM3Leon-7BRetrieval in Training=true, Responsible=true, Number of Retrieved Documents=1, Dataset Size=340M, Model Size=7B, Zero-shot=true, Samples per query=8, Selection=CLIP-based re-ranking2023.09 | 5.78 | — | |
| CM3Leon-760MRetrieval in Training=true, Responsible=true, Number of Retrieved Documents=2, Dataset Size=340M, Model Size=760M, Zero-shot=true, Samples per query=8, Selection=CLIP-based re-ranking2023.09 | 6.61 | — | |
| ImagenType=Gen. Only, Params=3B, Images=5000M+2024.10 | 6.61 | — | |
| PARTIRetrieval in Training=false, Responsible=false, Dataset Size=5B, Model Size=20B, Zero-shot=true2023.09 | 7.23 | — | |
| MUSERetrieval in Training=false, Responsible=false, Dataset Size=500M, Model Size=3B, Zero-shot=true2023.09 | 7.88 | — | |
| DreamLLMType=Und. and Gen. w/ pre-trained Diff., Params=7B, Images=N/A2024.10 | 8.76 | — | |
| Show-oType=Joint Prob. Models, Params=1.3B, Images=35M2024.10 | 9.24 | — | |
| UniDiffuserParam=0.9B2024.12 | 9.71 | 30.93 | |
| DALL-E2Type=Gen. Only, Params=6.5B, Images=650M2024.10 | 10.39 | — | |
| CM3Leon-7BRetrieval in Training=true, Responsible=true, Number of Retrieved Documents=0, Dataset Size=340M, Model Size=7B, Zero-shot=true, Samples per query=8, Selection=CLIP-based re-ranking2023.09 | 10.82 | — | |
| SDv1.5Param=0.9B2024.12 | 11.12 | 30.63 | |
| CoDiParam=4.3B2024.12 | 11.26 | 30.69 | |
| CODIType=Und. and Gen. w/ pre-trained Diff., Params=400M, Images=N/A2024.10 | 11.26 | — | |
| KNN-DiffusionRetrieval in Training=false, Responsible=false, Number of Retrieved Documents=10, Dataset Size=70M, Model Size=400M, Zero-shot=true2023.09 | 12.5 | — | |
| Stable DiffusionRetrieval in Training=false, Responsible=false, Number of Retrieved Documents=0, Dataset Size=400M, Model Size=800M, Zero-shot=true2023.09 | 12.6 | — | |
| LDMType=Gen. Only, Params=1.4B, Images=400M2024.10 | 12.64 | — | |
| SEED-XType=Und. and Gen. w/ pre-trained Diff., Params=17B, Images=N/A2024.10 | 12.68 | — | |
| UIO-2XXLParam=6.8B2024.12 | 13.39 | — | |
| OmniFlowParam=3.4B, Pretraining with high quality data=true2024.12 | 13.4 | 31.54 | |
| CM3Leon-350MRetrieval in Training=true, Responsible=true, Number of Retrieved Documents=2, Dataset Size=340M, Model Size=350M, Zero-shot=true, Samples per query=8, Selection=CLIP-based re-ranking2023.09 | 14.2 | — | |
| RA-CM3Retrieval in Training=true, Responsible=false, Number of Retrieved Documents=2, Dataset Size=150M, Model Size=2.7B, Zero-shot=true2023.09 | 15.7 | — | |
| SDXLParam=2.6B, Pretraining with high quality data=true2024.12 | 16.49 | 31.36 | |
| TransfusionType=Joint Prob. Models, Params=7B, Images=N/A2024.10 | 16.8 | — | |
| SingDiffusionBase Model=SD-2.0-base, Classifier Guidance=None2024.03 | 18.01 | 28.23 | |
| CCMBase Model=SDXL2024.12 | 20.47 | 31.73 | |
| SD3-MediumParam=2B, Pretraining with high quality data=true2024.12 | 20.94 | 30.65 | |
| SingDiffusionBase Model=SD-1.5, Classifier Guidance=None2024.03 | 21.09 | 27.71 | |
| PCMBase Model=SDXL2024.12 | 21.15 | 31.63 | |
| CCMBase Model=SD32024.12 | 21.49 | 31.41 | |
| MMAR-7BType=Joint Prob. Models, Params=7B, Images=145.2M2024.10 | 22.9 | — | |
| Stable Diffusion 2.0-baseBase Model=SD-2.0-base, Classifier Guidance=None2024.03 | 25.17 | 27.48 | |
| LCMBase Model=SD32024.12 | 25.44 | 31.27 | |
| DALL-EType=Gen. Only, Params=12B, Images=250M2024.10 | 27.5 | — | |
| PCMBase Model=SD32024.12 | 28.52 | 31.06 | |
| ChanmeleonType=Joint Prob. Models, Params=7B, Images=N/A2024.10 | 29.6 | — | |
| Hyper-SDBase Model=SDXL2024.12 | 30.87 | 31.3 | |
| Stable Diffusion 1.5Base Model=SD-1.5, Classifier Guidance=None2024.03 | 31.86 | 26.7 | |
| MMAR-0.5BType=Joint Prob. Models, Params=0.5B, Images=145.2M2024.10 | 36.6 | — | |
| OriginalBase Model=SD32024.12 | 86.84 | 26.18 |