Image Generation on Conditional ImageNet 128x128 (test)
1.94FIDSimple Diff
Evaluation results
| Method | Links | |
|---|---|---|
| Simple DiffType=Diff/FM2024.12 | 1.94 | |
| RINType=Diff/FM2024.12 | 2.75 | |
| ADM-GType=Diff/FM2024.12 | 2.97 | |
| CDMType=Diff/FM2024.12 | 3.52 | |
| TARFLOWType=NF, Patch Size (P)=4, Model Channel Size (Ch)=1024, Autoregressive Flow Blocks (T)=8, Attention Layers (K)=8, Input Noise (pε)=N(0, 0.15^2)2024.12 | 5.03 | |
| TARFLOWType=NF, Patch Size (P)=4, Model Channel Size (Ch)=1024, Autoregressive Flow Blocks (T)=8, Attention Layers (K)=8, Input Noise (pε)=N(0, 0.05^2)2024.12 | 5.29 | |
| BigGAN-deepType=GAN2024.12 | 5.7 | |
| BigGANType=GAN2024.12 | 8.7 |