Image Classification on ImageNet-1K (val) (Accuracy, GFLOPs, and Training Time)
80.1Top-1 AccuracyMCTF
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| MCTFBackbone=DeiT-S, Epochs=302025.05 | 80.1 | 2.6 | 39 | |
| DeiT-S (Baseline)Backbone=DeiT-S2025.05 | 79.8 | 4.6 | — | |
| ATMBackbone=DeiT-S, Epochs=30, Training-free=false2025.05 | 79.8 | 2.6 | 11 | |
| ATMBackbone=DeiT-S, Epochs=0, Training-free=true2025.05 | 79.7 | 3 | — | |
| BATBackbone=DeiT-S, Epochs=3002025.05 | 79.6 | 3 | — | |
| EViTBackbone=DeiT-S, Epochs=3002025.05 | 79.5 | 3 | 95 | |
| Evo-ViTBackbone=DeiT-S, Epochs=3002025.05 | 79.4 | 3 | 126 | |
| ToMeBackbone=DeiT-S, Epochs=3002025.05 | 79.4 | 2.7 | 87 | |
| DynamicViTBackbone=DeiT-S, Epochs=302025.05 | 79.3 | 2.9 | 43 | |
| IA-RED2Backbone=DeiT-S, Epochs=902025.05 | 79.1 | 3.2 | — | |
| A-ViTBackbone=DeiT-S, Epochs=1002025.05 | 78.6 | 3.6 | 65 |