Conditional Image Generation on ImageNet 256px 2012 (val)
1.12FIDSiT-XL/2+REPA-E
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| SiT-XL/2+REPA-ESpace=Joint-trained, Param=675M, Epoch=800, Guidance=CFG [19], NFE=250x22025.11 | 1.12 | 4.09 | 302.9 | 79 | 66 | |
| RAE-XLSpace=DINOv2 [60], Param=676M, Epoch=800, Guidance=AG [33], NFE=250x22025.11 | 1.41 | — | 309.4 | 80 | 63 | |
| SiT-XL/2+REPASpace=LDM [68], Param=675M, Epoch=800, Guidance=CFG [19], NFE=250x22025.11 | 1.42 | 4.7 | 305.7 | 80 | 65 | |
| MARSpace=LDM [68], Param=943M, Epoch=800, Guidance=CFG [19], NFE=64x100x22025.11 | 1.55 | — | 303.7 | 81 | 62 | |
| SiT-XL/2+REPA-ESpace=Joint-trained, Param=675M, Epoch=800, Guidance=None, NFE=2502025.11 | 1.69 | 4.17 | 219.3 | 77 | 67 | |
| RAE-XLSpace=DINOv2 [60], Param=676M, Epoch=800, Guidance=None, NFE=2502025.11 | 1.87 | — | 209.7 | 80 | 63 | |
| VARSpace=MS-VQVAE [84], Param=2.0B, Epoch=350, Guidance=CFG [19], NFE=10x22025.11 | 1.92 | — | 323.1 | 82 | 59 | |
| AF-XL/2 (4x deep, 112-layer)Space=LDM [68], Param=675M, Epoch=120, Guidance=CG[9]+DA[30], NFE=12025.11 | 1.94 | 4.54 | 292.2 | 79 | 56 | |
| SiT-XL/2+DisperseSpace=LDM [68], Param=675M, Epoch=1200, Guidance=CFG [19], NFE=500x22025.11 | 1.97 | — | — | — | — | |
| AF-XL/2Space=LDM [68], Param=675M, Epoch=145, Guidance=CG[9]+DA[30], NFE=42025.11 | 2.03 | 4.59 | 259.66 | 78 | 59 | |
| SiT-XL/2Space=LDM [68], Param=675M, Epoch=1400, Guidance=CFG [19], NFE=250x22025.11 | 2.06 | 4.49 | 277.5 | 83 | 59 | |
| AF-XL/2 (2x deep, 56-layer)Space=LDM [68], Param=675M, Epoch=95, Guidance=CG[9]+DA[30], NFE=12025.11 | 2.08 | 4.79 | 298.33 | 79 | 56 | |
| AF-XL/2Space=LDM [68], Param=675M, Epoch=95, Guidance=CG[9]+DA[30], NFE=22025.11 | 2.11 | 4.33 | 273.84 | 82 | 55 | |
| AlphaFlow-XL/2Space=LDM [68], Param=676M, Epoch=240, Guidance=CFG [19], NFE=22025.11 | 2.16 | — | — | — | — | |
| MeanFlow-XL/2Space=LDM [68], Param=676M, Epoch=1000, Guidance=CFG [19], NFE=22025.11 | 2.2 | — | — | — | — | |
| DiT-XL/2Space=LDM [68], Param=675M, Epoch=1400, Guidance=CFG [19], NFE=250x22025.11 | 2.27 | 4.6 | 278.24 | 83 | 57 | |
| StyleGAN-XLSpace=Pixel, Param=166M, Guidance=CG[9]+cGAN[58], NFE=12025.11 | 2.3 | 4.02 | 265.12 | 78 | 53 | |
| AF-XL/2Space=LDM [68], Param=673M, Epoch=90, Guidance=None, NFE=22025.11 | 2.36 | 4.35 | 235.77 | 81 | 52 | |
| AF-XL/2Space=LDM [68], Param=673M, Epoch=125, Guidance=CG[9]+DA[30], NFE=12025.11 | 2.38 | 4.87 | 284.18 | 81 | 52 | |
| AF-L/2Space=LDM [68], Param=457M, Epoch=120, Guidance=CG[9]+DA[30], NFE=12025.11 | 2.63 | 5.1 | 277.96 | 81 | 52 | |
| AlphaFlow-XL/2Space=LDM [68], Param=676M, Epoch=240, Guidance=CFG [19], NFE=12025.11 | 2.81 | — | — | — | — | |
| AF-M/2Space=LDM [68], Param=306M, Epoch=120, Guidance=CG[9]+DA[30], NFE=12025.11 | 2.82 | 5.2 | 279.12 | 81 | 50 | |
| GAT-XL/2+REPASpace=LDM [68], Param=602M, Epoch=40, Guidance=DA[30]+cGAN[58], NFE=12025.11 | 2.96 | — | — | — | — | |
| AF-B/2Space=LDM [68], Param=130M, Epoch=200, Guidance=CG[9]+DA[30], NFE=12025.11 | 3.05 | 5.32 | 269.18 | 81 | 51 | |
| TiM-XL/2Space=LDM [68], Param=664M, Epoch=300, Guidance=CFG [19], NFE=12025.11 | 3.26 | 4.37 | 210.33 | 75 | 59 | |
| MeanFlow-XL/2Space=LDM [68], Param=676M, Epoch=240, Guidance=CFG [19], NFE=12025.11 | 3.43 | — | — | — | — | |
| GigaGANSpace=Pixel, Param=569M, Epoch=480, Guidance=Match-loss [26], NFE=12025.11 | 3.45 | — | 225.52 | 84 | 61 | |
| TiM-XL/2Space=LDM [68], Param=664M, Epoch=300, Guidance=CFG [19], NFE=2x22025.11 | 3.61 | 6.74 | 151.79 | 74 | 59 | |
| MeanFlow-L/2Space=LDM [68], Param=459M, Epoch=240, Guidance=CFG [19], NFE=12025.11 | 3.84 | — | — | — | — | |
| AF-XL/2Space=LDM [68], Param=673M, Epoch=120, Guidance=None, NFE=12025.11 | 3.98 | 5.4 | 201.85 | 78 | 52 | |
| IMM-XL/2Space=LDM [68], Param=675M, Guidance=CFG [19], NFE=2x22025.11 | 3.99 | — | — | — | — | |
| AF-L/2Space=LDM [68], Param=457M, Epoch=110, Guidance=None, NFE=12025.11 | 4.36 | 5.39 | 186.21 | 77 | 53 | |
| ADM-GSpace=Pixel, Param=554M, Epoch=400, Guidance=CG [9], NFE=250x22025.11 | 4.59 | 5.25 | 186.7 | 82 | 52 | |
| VARSpace=MS-VQVAE [84], Param=310M, Epoch=350, Guidance=None, NFE=102025.11 | 4.95 | — | — | — | — | |
| MeanFlow-M/2Space=LDM [68], Param=308M, Epoch=240, Guidance=CFG [19], NFE=12025.11 | 5.01 | — | — | — | — | |
| AF-M/2Space=LDM [68], Param=306M, Epoch=110, Guidance=None, NFE=12025.11 | 5.21 | 5.6 | 178.48 | 75 | 54 | |
| AlphaFlow-B/2Space=LDM [68], Param=131M, Epoch=240, Guidance=CFG [19], NFE=12025.11 | 5.4 | — | — | — | — | |
| SiT-XL/2+REPASpace=LDM [68], Param=675M, Epoch=800, Guidance=None, NFE=2502025.11 | 5.9 | — | — | — | — | |
| AF-B/2Space=LDM [68], Param=130M, Epoch=170, Guidance=None, NFE=12025.11 | 6.07 | 5.31 | 169.51 | 72 | 49 | |
| MeanFlow-B/2Space=LDM [68], Param=131M, Epoch=240, Guidance=CFG [19], NFE=12025.11 | 6.17 | — | — | — | — | |
| MaskGITSpace=VQGAN [10], Param=227M, Epoch=300, Guidance=None, NFE=82025.11 | 6.18 | — | 182.1 | 80 | 51 | |
| BigGANSpace=Pixel, Param=112M, Guidance=cGAN [58], NFE=12025.11 | 6.95 | 7.36 | 171.4 | 87 | 28 | |
| TiM-XL/2Space=LDM [68], Param=664M, Epoch=300, Guidance=None, NFE=12025.11 | 7.11 | 4.97 | 140.39 | 71 | 63 | |
| SiT-XL/2+DisperseSpace=LDM [68], Param=675M, Epoch=1200, Guidance=None, NFE=5002025.11 | 7.43 | — | — | — | — | |
| IMM-XL/2Space=LDM [68], Param=675M, Guidance=CFG [19], NFE=1x22025.11 | 7.77 | — | — | — | — | |
| Shortcut-XL/2Space=LDM [68], Param=675M, Epoch=250, Guidance=CFG [19], NFE=42025.11 | 7.8 | — | — | — | — | |
| SiT-XL/2Space=LDM [68], Param=675M, Epoch=1400, Guidance=None, NFE=2502025.11 | 8.3 | — | — | — | — | |
| DiT-XL/2Space=LDM [68], Param=675M, Epoch=1400, Guidance=None, NFE=2502025.11 | 9.62 | 6.85 | 121.5 | 67 | 67 | |
| Shortcut-XL/2Space=LDM [68], Param=675M, Epoch=250, Guidance=CFG [19], NFE=12025.11 | 10.6 | — | — | — | — | |
| ADMSpace=Pixel, Param=554M, Epoch=400, Guidance=None, NFE=2502025.11 | 10.94 | 6.02 | 100.98 | 69 | 63 |