Image Classification on ImageNet (Top-1 Accuracy, Throughput, and Speedup)
79.8Top-1 AccuracyDeiT
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| DeiTLayers=None, Retention Ratio (Ret.)=1.002026.05 | 79.8 | 12.36 | 3.12 | 1 | 1 | |
| S2S-MambaLayers=9–11, Retention Ratio (Ret.)=1.002026.05 | 79.3 | 10.46 | 3.81 | 0.85 | 1.22 | |
| A-ViTLayers=None, Retention Ratio (Ret.)=0.782026.05 | 78.6 | 13.51 | 4.08 | 1.09 | 1.31 | |
| S2S-MambaLayers=9–11, Retention Ratio (Ret.)=0.782026.05 | 78.6 | 13.49 | 5.11 | 1.09 | 1.64 | |
| S2S-MambaLayers=9–11, Retention Ratio (Ret.)=0.702026.05 | 78.4 | 14.45 | 5.32 | 1.17 | 1.71 | |
| A-ViTLayers=None, Retention Ratio (Ret.)=0.702026.05 | 78.2 | 15.03 | 4.71 | 1.22 | 1.51 |