Class-conditional Image Generation on ImageNet 512
2.35FIDEPG-L/32
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| EPG-L/32Model Type=Models in Pixel Space, NFE=75, Epochs=800, #Params=540M, GFLOPs=1132025.10 | 2.35 | 0.82 | 0.57 | — | — | 295.4 | |
| EPG-L/32Model Type=Models in Pixel Space, NFE=75, Epochs=600, #Params=540M, GFLOPs=1132025.10 | 2.43 | 0.82 | 0.57 | — | — | 289.7 | |
| MaskDiTModel Type=Models in Latent Space, NFE=79×2, Epochs=800, #Params=84M + 675M, GFLOPs=1260 + 524.62025.10 | 2.5 | 0.83 | 0.56 | — | — | 256.7 | |
| SiT-XL/2Model Type=Models in Latent Space, NFE=250×2, Epochs=600, #Params=84M + 675M, GFLOPs=1260 + 524.62025.10 | 2.62 | 0.84 | 0.57 | — | — | 252.2 | |
| VDM++Model Type=Models in Pixel Space, NFE=250×2, Epochs=800, #Params=2.46B2025.10 | 2.65 | — | — | — | — | 278.1 | |
| PixelNerdModel Type=Models in Pixel Space, NFE=100×2, Epochs=320, #Params=700M, GFLOPs=5832025.10 | 2.84 | 0.8 | 0.59 | — | — | 245.6 | |
| SiDModel Type=Models in Pixel Space, NFE=250×2, Epochs=800, #Params=2.46B2025.10 | 3.02 | — | — | — | — | 248.7 | |
| DiT-XL/2Model Type=Models in Latent Space, NFE=250×2, Epochs=600, #Params=84M + 675M, GFLOPs=1260 + 524.62025.10 | 3.04 | 0.84 | 0.54 | — | — | 240.8 | |
| ADMModel Type=Models in Pixel Space, NFE=500, Epochs=1081, #Params=731M, GFLOPs=28132025.10 | 3.85 | 0.84 | 0.53 | — | — | 221.7 | |
| RINModel Type=Models in Pixel Space, NFE=1000, Epochs=800, #Params=320M, GFLOPs=4152025.10 | 3.95 | — | — | — | — | 216 | |
| U-ViT-H/4Model Type=Models in Latent Space, NFE=50×2, Epochs=400, #Params=84M + 501M, GFLOPs=1260 + 133.02025.10 | 4.05 | 0.84 | 0.48 | — | — | 263.8 | |
| CADSSampler=CADS, Guidance Scale (wCFG)=52023.10 | 9.81 | 0.82 | 0.52 | 0.09 | 7.75 | — | |
| DDPMSampler=DDPM, Guidance Scale (wCFG)=52023.10 | 23.1 | 0.82 | 0.28 | 0.2 | 4.88 | — |