Image Generation on ImageNet 50K
2.27FIDDiT-XL/2
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| DiT-XL/2#Params=675M, #Step=250, Time (s / img)=2.592, Rel. Time=452025.06 | 2.27 | 278.2 | 83 | 57 | |
| DiT-XL/2#Params=675M, #Step=250, Time (s / img)=2.592, Rel. Time=452025.06 | 2.27 | 278.2 | 0.83 | 0.57 | |
| LDM-4-G#Params=400M, #Step=2502025.06 | 3.6 | 247.7 | — | — | |
| LDM-4-G#Params=400M, #Step=2502025.06 | 3.6 | 247.7 | — | — | |
| ViT-VQGAN#Params=1.7B, #Step=1024, Time (s / img)=1.382, Rel. Time=242025.06 | 4.17 | 175.1 | — | — | |
| ViT-VQGAN#Params=1.7B, #Step=1024, Time (s / img)=1.382, Rel. Time=242025.06 | 4.17 | 175.1 | — | — | |
| CDM#Step=81002025.06 | 4.88 | 158.7 | — | — | |
| CDM#Step=81002025.06 | 4.88 | 158.7 | — | — | |
| DiT-L/2#Params=458M, #Step=250, Time (s / img)=1.786, Rel. Time=312025.06 | 5.02 | 167.2 | 75 | 57 | |
| DiT-L/2#Params=458M, #Step=250, Time (s / img)=1.786, Rel. Time=312025.06 | 5.02 | 167.2 | 0.75 | 0.57 | |
| LaTtE-Flow#Params=500M, #Step=40, Time (s / img)=0.052, Rel. Time=12025.06 | 5.79 | 213.1 | 78 | 69 | |
| LaTtE-Flow#Params=500M, #Step=40, Time (s / img)=0.052, Rel. Time=12025.06 | 5.79 | 213.1 | 0.78 | 0.69 | |
| LaTtE-Flow Blend#Params=500M, #Step=40, Time (s / img)=0.061, Rel. Time=12025.06 | 6.03 | 193.9 | 77 | 68 | |
| Vanilla Blend#Params=2.0B, #Step=40, Time (s / img)=0.185, Rel. Time=42025.06 | 6.12 | 193.7 | 78 | 69 | |
| MaskGIT#Params=227M, #Step=8, Time (s / img)=0.029, Rel. Time=0.52025.06 | 6.18 | 182.1 | 80 | 51 | |
| MaskGIT#Params=227M, #Step=8, Time (s / img)=0.029, Rel. Time=0.52025.06 | 6.18 | 182.1 | 0.8 | 0.51 | |
| Vanilla#Params=2.0B, #Step=40, Time (s / img)=0.158, Rel. Time=32025.06 | 6.33 | 192.4 | 80 | 67 | |
| Vanilla#Params=2.0B, #Step=40, Time (s / img)=0.158, Rel. Time=32025.06 | 6.33 | 192.4 | 0.8 | 0.67 | |
| MAGE#Params=230M2025.06 | 6.93 | 195.8 | — | — | |
| MAGE#Params=230M2025.06 | 6.93 | 195.8 | — | — | |
| RQTran.#Params=3.8B, #Step=68, Time (s / img)=1.210, Rel. Time=212025.06 | 7.55 | 134 | — | — | |
| RQTran.#Params=3.8B, #Step=68, Time (s / img)=1.210, Rel. Time=212025.06 | 7.55 | 134 | — | — | |
| ADM#Params=554M, #Step=250, Time (s / img)=9.677, Rel. Time=1682025.06 | 10.94 | 101 | 69 | 63 | |
| ADM#Params=554M, #Step=250, Time (s / img)=9.677, Rel. Time=1682025.06 | 10.94 | 101 | 0.69 | 0.63 | |
| VQGAN#Params=1.4B, #Step=256, Time (s / img)=1.382, Rel. Time=242025.06 | 15.78 | 74.3 | — | — | |
| VQGAN#Params=1.4B, #Step=256, Time (s / img)=1.382, Rel. Time=242025.06 | 15.78 | 74.3 | — | — | |
| VQGAN#Params=227M, #Step=256, Time (s / img)=1.094, Rel. Time=192025.06 | 18.65 | 80.4 | 78 | 26 | |
| VQGAN (†)#Params=227M, #Step=256, Time (s / img)=1.094, Rel. Time=192025.06 | 18.65 | 80.4 | 0.78 | 0.26 | |
| Janus Pro#Params=1.5B, #Step=576, Time (s / img)=0.311, Rel. Time=62025.06 | 23.68 | 105.2 | 58 | 49 | |
| Janus Pro#Params=1.5B, #Step=576, Time (s / img)=0.311, Rel. Time=62025.06 | 23.68 | 105.2 | 0.58 | 0.49 | |
| VQVAE-2#Params=13.5B, #Step=51202025.06 | 31.11 | 45 | 36 | 57 | |
| VQVAE-2#Params=13.5B, #Step=51202025.06 | 31.11 | 45 | 0.36 | 0.57 | |
| Show-o#Params=1.3B, #Step=50, Time (s / img)=2.493, Rel. Time=482025.06 | 31.26 | 98.7 | 55 | 69 | |
| Show-o#Params=1.3B, #Step=50, Time (s / img)=2.493, Rel. Time=482025.06 | 31.26 | 98.7 | 0.55 | 0.69 |