Image Generation on ImageNet 32x32 (test)
3.72FIDVDM
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| VDMModel Category=Likelihood-based2021.06 | 3.72 | — | — | — | — | — | |
| Teacher2024.09 | 4.67 | — | 13.63 | — | — | 4.03 | |
| ScoreFlowVariant=VP, FID, Model Category=Likelihood-based2021.06 | 8.34 | 3.84 | — | — | — | — | |
| DDPM++ (VP, NLL) + STSoft Truncation=true, Backbone=DDPM++, Variant=VP, Target=NLL2021.06 | 8.42 | 3.85 | 11.82 | — | — | — | |
| ScoreFlowVariant=deep, sub-VP, NLL, Model Category=Likelihood-based2021.06 | 10.18 | 3.76 | — | — | — | — | |
| Data-Based Training2024.09 | 10.56 | — | 9.99 | — | — | 5.24 | |
| Dynamic Iterative Distillation (DKDM)2024.09 | 11.33 | — | 10.5 | — | — | 4.8 | |
| Data-Limited TrainingReal Data Percentage=20%2024.09 | 12.62 | — | 10.48 | — | — | 9.7 | |
| Data-Free TrainingReal Data Percentage=0%2024.09 | 13.2 | — | 10.47 | — | — | 9.56 | |
| Data-Limited TrainingReal Data Percentage=15%2024.09 | 13.5 | — | 10.43 | — | — | 10.76 | |
| Data-Limited TrainingReal Data Percentage=5%2024.09 | 13.63 | — | 10.48 | — | — | 10.62 | |
| Data-Limited TrainingReal Data Percentage=10%2024.09 | 14.23 | — | 10.39 | — | — | 12.63 | |
| QC-NCSN++NFE=1000, Sampler=PC sampler2022.09 | 19.62 | — | 9.94 | 61 | 52 | — | |
| U-NCSN++NFE=1000, Sampler=PC sampler2022.09 | 19.82 | — | 9.89 | 60 | 52 | — | |
| C-NCSN++NFE=1000, Sampler=PC sampler2022.09 | 28.97 | — | 8.58 | 61 | 45 | — | |
| DenseFlow-74-10Model Category=Likelihood-based2021.06 | — | 3.63 | — | — | — | — | |
| NDMintegration=Ordinary Differential Equation (ODE)2023.10 | — | 3.55 | — | — | — | — | |
| Score Flow2023.10 | — | 3.76 | — | — | — | — | |
| VDM2023.10 | — | 3.72 | — | — | — | — |