Image Classification on ImageNet-1K (val) (Top-1 Accuracy under Occlusion)
80.7Top-1 Acc (0% Occ)MergeMix
Evaluation results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| MergeMixBackbone=DeiT-Small, Training Epochs=300, Inference with token merging (ToMe)=false2025.10 | 80.7 | 80.38 | 79.95 | 78.97 | 78.08 | 76.21 | 74.08 | 70.71 | 65.57 | 51.98 | |
| TransMixBackbone=DeiT-Small, Training Epochs=300, Inference with token merging (ToMe)=false2025.10 | 80.36 | 79.47 | 78.24 | 77 | 75.4 | 73.31 | 70.17 | 65.78 | 57.79 | 42.44 | |
| DeiTBackbone=DeiT-Small, Training Epochs=300, Inference with token merging (ToMe)=false2025.10 | 80.27 | 79.03 | 77.92 | 76.39 | 74.65 | 72.25 | 68.31 | 63.53 | 54.57 | 38.81 | |
| CutMixBackbone=DeiT-Small, Training Epochs=300, Inference with token merging (ToMe)=false2025.10 | 80.16 | 79.23 | 78.2 | 76.87 | 75.21 | 72.92 | 69.24 | 64.63 | 55.95 | 39.69 | |
| MergeMixBackbone=DeiT-Small, Training Epochs=300, Inference with token merging (ToMe)=true2025.10 | 79.67 | 79.57 | 79.01 | 78.6 | 77.68 | 76.11 | 73.99 | 70.56 | 64.84 | 51.8 | |
| SMMixBackbone=DeiT-Small, Training Epochs=300, Inference with token merging (ToMe)=false2025.10 | 79.32 | 78.56 | 77.7 | 76.17 | 73.76 | 70.25 | 64.95 | 58.94 | 48.69 | 34.37 | |
| MixProBackbone=DeiT-Small, Training Epochs=300, Inference with token merging (ToMe)=false2025.10 | 79.25 | 78.83 | 78.01 | 77.24 | 76.02 | 74.26 | 71.48 | 67.44 | 58.49 | 39.47 | |
| MixUpBackbone=DeiT-Small, Training Epochs=300, Inference with token merging (ToMe)=false2025.10 | 78.74 | 78.17 | 77.11 | 76.21 | 74.31 | 71.85 | 68.13 | 62.4 | 51.45 | 33.08 | |
| VanillaBackbone=DeiT-Small, Training Epochs=300, Inference with token merging (ToMe)=false2025.10 | 75.46 | 74.03 | 72.85 | 71.36 | 69.91 | 67.94 | 64.71 | 60.42 | 51.65 | 34.08 |