Image Generation on ImageNet 128x128 (val)
1.26FIDSiD2, Flop Heavy
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SiD2, Flop HeavyNFE=5122024.10 | 1.26 | — | |
| PaGoDANFE=12024.10 | 1.48 | — | |
| DisCo-DiffSampler=ODE sampler, Correction=VDM++ correction2024.07 | 1.73 | 414 | |
| DisCo-DiffNFE=4142024.10 | 1.73 | — | |
| VDM+++NFE=2502024.10 | 1.75 | — | |
| StyleGAN-XLNFE=12024.10 | 1.81 | — | |
| VDM++Sampler=DDPM sampler2024.07 | 1.88 | 512 | |
| SD, U-ViT-LNFE=5122024.10 | 1.94 | — | |
| DisCo-DiffSampler=ODE sampler2024.07 | 1.98 | 114 | |
| VDM++Sampler=ODE sampler2024.07 | 2.29 | 115 | |
| RIN2024.07 | 2.75 | 1,000 | |
| RINNFE=10002024.10 | 2.75 | — | |
| ADM-GGuidance=Classifier guidance2024.07 | 2.97 | 250 | |
| CDMResolution stages=32, 64, 1282024.07 | 3.52 | 8,100 | |
| ADM2024.07 | 5.91 | 250 | |
| NSVQtokens=256, Codebook size=65,536, Latent dim=1282026.06 | 16.62 | — | |
| VQGAN-LCtokens=256, Codebook size=65,536, Latent dim=1282026.06 | 17.98 | — | |
| VQGANtokens=256, Codebook size=65,536, Latent dim=1282026.06 | 19.57 | — | |
| SimVQtokens=256, Codebook size=65,536, Latent dim=1282026.06 | 20.11 | — |