Dataset Condensation on ImageNet 1K 2015 (train val)
3.56FIDGADC
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| GADCArchitecture=SiT-L/2, Data Budget=0.8% (10K), Eval. Samples=50K, Training Iterations=100K, Resolution=256x2562026.06 | 3.56 | 415.4 | 78 | 27 | |
| GADCArchitecture=SiT-L/2, Data Budget=0.8% (10K), Eval. Samples=10K, Training Iterations=100K, Resolution=256x2562026.06 | 5.91 | 423.2 | 78 | 70 | |
| CCSArchitecture=SiT-L/2, Data Budget=0.8% (10K), Eval. Samples=50K, Training Iterations=100K, Resolution=256x2562026.06 | 6.54 | 373.1 | 78 | 19 | |
| CCSArchitecture=SiT-L/2, Data Budget=0.8% (10K), Eval. Samples=10K, Training Iterations=100K, Resolution=256x2562026.06 | 9.52 | 377 | 78 | 58 | |
| IGDArchitecture=SiT-L/2, Data Budget=0.8% (10K), Eval. Samples=50K, Training Iterations=100K, Resolution=256x2562026.06 | 13.2 | 427 | 86 | 12 | |
| IGDArchitecture=SiT-L/2, Data Budget=0.8% (10K), Eval. Samples=10K, Training Iterations=100K, Resolution=256x2562026.06 | 16.1 | 427.5 | 86 | 43 | |
| RDEDArchitecture=SiT-L/2, Data Budget=0.8% (10K), Eval. Samples=50K, Training Iterations=100K, Resolution=256x2562026.06 | 22.7 | 483.1 | 82 | 6 | |
| RDEDArchitecture=SiT-L/2, Data Budget=0.8% (10K), Eval. Samples=10K, Training Iterations=100K, Resolution=256x2562026.06 | 26 | 490.3 | 82 | 26 | |
| EDCArchitecture=SiT-L/2, Data Budget=0.8% (10K), Eval. Samples=50K, Training Iterations=100K, Resolution=256x2562026.06 | 26.7 | 169.9 | 54 | 22 | |
| EDCArchitecture=SiT-L/2, Data Budget=0.8% (10K), Eval. Samples=10K, Training Iterations=100K, Resolution=256x2562026.06 | 29.3 | 171.8 | 55 | 64 |