Image Classification on ImageNet-21K (val)
46.4Top-1 AccuracyEfficientNetV2-L with Poly-1 loss
Evaluation results
| Method | Links | |
|---|---|---|
| EfficientNetV2-L with Poly-1 lossBackbone=EfficientNetV2-L, Loss Function=L_Poly-1, epsilon_1=22022.04 | 46.4 | |
| EfficientNetV2-L with Cross-entropy lossBackbone=EfficientNetV2-L, Loss Function=L_CE2022.04 | 45.8 | |
| Ours5×IPC=20, Evaluation model=RegNet-Y-8GF2026.02 | 37.4 | |
| Ours5×IPC=20, Evaluation model=ResNet-1012026.02 | 37.3 | |
| Ours5×IPC=20, Evaluation model=ConvNeXt-Tiny2026.02 | 37.2 | |
| Ours5×IPC=20, Evaluation model=ResNet-502026.02 | 36.9 | |
| D3SIPC=20, Evaluation model=RegNet-Y-8GF2026.02 | 36.4 | |
| CDAIPC=20, Evaluation model=ConvNeXt-Tiny2026.02 | 36.3 | |
| CDAIPC=20, Evaluation model=ResNet-1012026.02 | 36.1 | |
| CDAIPC=20, Evaluation model=RegNet-Y-8GF2026.02 | 36.1 | |
| E2DIPC=20, Evaluation Model=ResNet-18, Acceleration Factor=5x2026.02 | 36 | |
| D3SIPC=20, Evaluation model=ResNet-1012026.02 | 36 | |
| Ours5×IPC=20, Evaluation model=ResNet-182026.02 | 36 | |
| E2D (5x acceleration)IPC=20, Evaluation model=ResNet-18, Acceleration=5x2026.02 | 36 | |
| D3SIPC=20, Evaluation model=ResNet-502026.02 | 35.4 | |
| CDAIPC=20, Evaluation model=ResNet-502026.02 | 35.3 | |
| SRe2LIPC=20, Evaluation model=ConvNeXt-Tiny2026.02 | 34.9 | |
| SRe2LIPC=20, Evaluation model=RegNet-Y-8GF2026.02 | 34.2 | |
| E2D (optimization-free)IPC=20, Evaluation Model=ResNet-18, Optimization Mode=free2026.02 | 33.9 | |
| E2D (optimization-free)IPC=20, Evaluation model=ResNet-18, Optimization-free=true2026.02 | 33.9 | |
| SRe2LIPC=20, Evaluation model=ResNet-1012026.02 | 33.2 | |
| Ours5×IPC=20, Evaluation model=DenseNet-1212026.02 | 33.1 | |
| E2DIPC=10, Evaluation Model=ResNet-18, Acceleration Factor=5x2026.02 | 32.1 | |
| E2D (5x acceleration)IPC=10, Evaluation model=ResNet-18, Acceleration=5x2026.02 | 32.1 | |
| D3SIPC=20, Evaluation model=DenseNet-1212026.02 | 31.9 | |
| SRe2LIPC=20, Evaluation model=ResNet-502026.02 | 31.2 | |
| E2D (optimization-free)IPC=10, Evaluation Model=ResNet-18, Optimization Mode=free2026.02 | 28.8 | |
| E2D (optimization-free)IPC=10, Evaluation model=ResNet-18, Optimization-free=true2026.02 | 28.8 | |
| CDAIPC=20, Evaluation model=DenseNet-1212026.02 | 28.6 | |
| D3SIPC=20, Evaluation Model=ResNet-182026.02 | 28.5 | |
| D3SIPC=20, Evaluation model=ResNet-182026.02 | 28.5 | |
| D3SIPC=20, Evaluation model=ResNet-182026.02 | 28.5 | |
| D3SIPC=10, Evaluation Model=ResNet-182026.02 | 26.9 | |
| D3SIPC=10, Evaluation model=ResNet-182026.02 | 26.9 | |
| CDAIPC=20, Evaluation Model=ResNet-182026.02 | 26.4 | |
| CDAIPC=20, Evaluation model=ResNet-182026.02 | 26.4 | |
| CDAIPC=20, Evaluation model=ResNet-182026.02 | 26.4 | |
| SRe2LIPC=20, Evaluation model=DenseNet-1212026.02 | 24.6 | |
| CDAIPC=10, Evaluation Model=ResNet-182026.02 | 22.6 | |
| CDAIPC=10, Evaluation model=ResNet-182026.02 | 22.6 | |
| SRe2LIPC=20, Evaluation Model=ResNet-182026.02 | 21.8 | |
| SRe2LIPC=20, Evaluation model=ResNet-182026.02 | 21.8 | |
| SRe2LIPC=20, Evaluation model=ResNet-182026.02 | 21.8 | |
| SRe2LIPC=10, Evaluation Model=ResNet-182026.02 | 18.5 | |
| SRe2LIPC=10, Evaluation model=ResNet-182026.02 | 18.5 |