Class-to-Image Generation on ImageNet 256x256 (train test)
1.63FIDPixelU
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| PixelUVariant=H/16, GFLOPs=136.9, CFG=true2026.06 | 1.63 | 305.88 | |
| JiTVariant=G/16, GFLOPs=383, CFG=true2026.06 | 1.82 | 292.6 | |
| JiTVariant=H/16, GFLOPs=182, CFG=true2026.06 | 1.86 | 303.4 | |
| PixelUVariant=L/16, GFLOPs=73.1, CFG=true2026.06 | 2.17 | 295.88 | |
| JiTVariant=L/16, GFLOPs=88, CFG=true2026.06 | 2.36 | 298.5 | |
| PixelUVariant=B/16, GFLOPs=21.8, CFG=true2026.06 | 3.4 | 276.01 | |
| JiTVariant=B/16, GFLOPs=25, CFG=true2026.06 | 3.66 | 275.1 |