Text-to-Image Generation on COCO 30k subset 2014 (val)
2.47FIDRandom Train Images
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Random Train ImagesSetting=fine-tuned2022.11 | 2.47 | — | |
| PartiSetting=fine-tuned2022.11 | 3.22 | — | |
| ReCoSetting=fine-tuned, Guidance Scale=1.52022.11 | 5.18 | — | |
| ReCo position WordSetting=fine-tuned, Guidance Scale=1.52022.11 | 5.98 | — | |
| Retrieval BaselineSetting=fine-tuned2022.11 | 6.82 | — | |
| eDiff-IResolution=256x256, Category=Proprietary Technologies2023.10 | 6.95 | — | |
| eDiff-IResolution=256, Latency=32.0s, Family=Original, unaccelerated2024.05 | 6.95 | — | |
| ReCo Image Descr.Setting=fine-tuned, Guidance Scale=1.52022.11 | 6.98 | — | |
| SDv1.5Resolution=512, Latency=10.25s, Family=Teacher, Steps=200, Sampler=SDE, cfg=22024.05 | 7.21 | — | |
| ImagenResolution=256x256, Category=Proprietary Technologies2023.10 | 7.27 | — | |
| ImagenResolution=256, Latency=9.1s, Family=Original, unaccelerated2024.05 | 7.27 | — | |
| Make-A-SceneSetting=fine-tuned2022.11 | 7.55 | — | |
| KandinskyResolution=256x256, Category=Open Sourced Technologies2023.10 | 8.03 | — | |
| Parti-3BResolution=256, Latency=6.4s, Family=Original, unaccelerated2024.05 | 8.1 | — | |
| LAFITESetting=fine-tuned2022.11 | 8.12 | — | |
| DMD2Resolution=512, Latency=0.09s, Family=Accelerated diffusion2024.05 | 8.35 | — | |
| Stable Diffusion 2.1Resolution=256x256, Category=Open Sourced Technologies2023.10 | 8.59 | — | |
| SDv1.5Resolution=512, Latency=2.59s, Family=Teacher, Steps=50, Sampler=ODE, cfg=32024.05 | 8.59 | — | |
| GigaGANResolution=256x256, Category=Proprietary Technologies2023.10 | 9.09 | — | |
| GigaGANResolution=512, Latency=0.13s, Family=GANs2024.05 | 9.09 | — | |
| XMC-GANSetting=fine-tuned2022.11 | 9.33 | — | |
| SLAM (4 step)Resolution=512, Latency=0.19s, Family=Accelerated diffusion2024.05 | 10.06 | — | |
| DALL-E 2Resolution=256x256, Category=Proprietary Technologies2023.10 | 10.39 | — | |
| DALL·E 2Resolution=256, Family=Original, unaccelerated2024.05 | 10.39 | — | |
| Parti-750MResolution=256, Family=Original, unaccelerated2024.05 | 10.71 | — | |
| DMDResolution=512, Latency=0.09s, Family=Accelerated diffusion2024.05 | 11.49 | — | |
| Make-A-SceneResolution=256, Latency=25.0s, Family=Original, unaccelerated2024.05 | 11.84 | — | |
| GLIDEResolution=256x256, Category=Open Sourced Technologies2023.10 | 12.24 | — | |
| GLIDEResolution=256, Latency=15.0s, Family=Original, unaccelerated2024.05 | 12.24 | — | |
| LDMResolution=256, Latency=3.7s, Family=Original, unaccelerated2024.05 | 12.63 | — | |
| UFOGenResolution=512, Latency=0.09s, Family=Accelerated diffusion2024.05 | 12.78 | — | |
| InstaFlow-0.9BResolution=512, Latency=0.09s, Family=Accelerated diffusion2024.05 | 13.1 | — | |
| StyleGAN-TResolution=512, Latency=0.10s, Family=GANs2024.05 | 13.9 | — | |
| HiPAResolution=512, Latency=0.09s, Family=Accelerated diffusion2024.05 | 13.91 | — | |
| IF*Resolution=256x256, Category=Open Sourced Technologies2023.10 | 15.1 | — | |
| Kandinsky 1.0Resolution=256x256, Category=Open Sourced Technologies2023.10 | 15.4 | — | |
| SwiftBrushResolution=512, Latency=0.09s, Family=Accelerated diffusion2024.05 | 16.67 | — | |
| CogView2Setting=fine-tuned2022.11 | 17.7 | — | |
| DALL-EResolution=256x256, Category=Proprietary Technologies2023.10 | 17.89 | — | |
| UniPC (4 step)Resolution=512, Latency=0.26s, Family=Accelerated diffusion2024.05 | 19.57 | — | |
| ruDALL-E MalevichResolution=256x256, Category=Open Sourced Technologies2023.10 | 20 | — | |
| GLIGENResolution=256x256, Category=Open Sourced Technologies2023.10 | 21.04 | — | |
| DPM++ (4 step)Resolution=512, Latency=0.26s, Family=Accelerated diffusion2024.05 | 22.36 | — | |
| LCM-LoRA (4 step)Resolution=512, Latency=0.19s, Family=Accelerated diffusion2024.05 | 23.62 | — | |
| LAFITEResolution=256, Latency=0.02s, Family=GANs2024.05 | 26.94 | — | |
| DALL·EResolution=256, Family=Original, unaccelerated2024.05 | 27.5 | — | |
| DALLEType=AR, Number of Parameters=12B2023.11 | — | 27.5 | |
| DALLE 2Type=Diff, Number of Parameters=5.5B2023.11 | — | 10.39 | |
| eDiff-IType=Diff, Inference Time (s)=32, Number of Parameters=9B2023.11 | — | 6.95 | |
| GigaGANType=GAN, Inference Time (s)=0.13, Number of Parameters=1B2023.11 | — | 9.09 | |
| GLIDEType=Diff, Inference Time (s)=15, Number of Parameters=5B2023.11 | — | 12.24 | |
| ImagenType=Diff, Inference Time (s)=9.1, Number of Parameters=3B2023.11 | — | 7.27 | |
| InstaFlowInference Time (s)=0.09, Number of Parameters=0.9B2023.11 | — | 13.1 | |
| LAFITEType=GAN, Inference Time (s)=0.02, Number of Parameters=75M2023.11 | — | 26.94 | |
| Make-A-SceneType=AR, Inference Time (s)=252023.11 | — | 11.84 | |
| Muse-3BInference Time (s)=1.3, Number of Parameters=3B2023.11 | — | 7.88 | |
| Parti-20BType=AR, Number of Parameters=20B2023.11 | — | 7.23 | |
| Stable DiffusionType=Diff, Inference Time (s)=2.9, Number of Parameters=0.9B2023.11 | — | 9.62 | |
| StyleGAN-TType=GAN, Inference Time (s)=0.1, Number of Parameters=1B2023.11 | — | 13.9 | |
| UFOGenInference Time (s)=0.09, Number of Parameters=0.9B2023.11 | — | 12.78 |