Conditional Image Generation on ImageNet 256x256 2012 (400K iterations)
3.89FIDDiC-XL
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| DiC-XLcfg=1.5, FLOPs (G)=116.1 (57.2)2024.12 | 3.89 | 224.2 | |
| U-ViT-XLcfg=1.5, FLOPs (G)=113.02024.12 | 5.66 | 170.62 | |
| DiT-XL/2cfg=1.5, FLOPs (G)=118.62024.12 | 6.24 | 150.1 | |
| DiC-XLFLOPs (G)=116.1, Training Iterations=400K2024.12 | 13.11 | 100.15 | |
| DiT-XL/2FLOPs (G)=118.6, Training Iterations=400K2024.12 | 20.05 | 66.74 | |
| DiC-BFLOPs (G)=23.5, Training Iterations=400K2024.12 | 32.33 | 48.72 | |
| DiT-B/2FLOPs (G)=23, Training Iterations=400K2024.12 | 42.84 | 33.66 | |
| DiC-SFLOPs (G)=5.9, Training Iterations=400K2024.12 | 58.68 | 25.82 | |
| DiT-S/2FLOPs (G)=6.1, Training Iterations=400K2024.12 | 67.4 | 20.44 |