Image Classification on CIFAR-100 (Accuracy & Latency)
80.58Top-1 AccuracyOA-CutMix/rel
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| OA-CutMix/relBackbone=DeiT-Ti, Epochs=400, Batch size=128, Training Protocol=from scratch2026.06 | 80.58 | 75.92 | — | |
| OA-CutMix/relBackbone=DeiT-S, Epochs=400, Batch size=128, Training Protocol=from scratch2026.06 | 80.42 | 75.92 | 1.03 | |
| CutMix + decoupledBackbone=DeiT-S, Epochs=400, Batch size=128, Training Protocol=from scratch2026.06 | 80.32 | 74.86 | 0.67 | |
| CutMix + decoupledBackbone=DeiT-Ti, Epochs=400, Batch size=128, Training Protocol=from scratch2026.06 | 79.93 | 74.86 | — | |
| Attentive CutMixBackbone=DeiT-S, Epochs=400, Batch size=128, Training Protocol=from scratch2026.06 | 79.92 | 73.56 | 56.1 | |
| CutMixBackbone=DeiT-Ti, Epochs=400, Batch size=128, Training Protocol=from scratch2026.06 | 79.83 | 74.02 | — | |
| SAMixBackbone=DeiT-Ti, Epochs=400, Batch size=128, Training Protocol=from scratch2026.06 | 79.72 | 72.65 | — | |
| Attentive CutMixBackbone=DeiT-Ti, Epochs=400, Batch size=128, Training Protocol=from scratch2026.06 | 79.49 | 73.56 | — | |
| SAMixBackbone=DeiT-S, Epochs=400, Batch size=128, Training Protocol=from scratch2026.06 | 79.48 | 72.65 | 88.27 | |
| CutMixBackbone=DeiT-S, Epochs=400, Batch size=128, Training Protocol=from scratch2026.06 | 79.14 | 74.02 | 0.67 | |
| TokenMixBackbone=DeiT-S, Epochs=400, Batch size=128, Training Protocol=from scratch2026.06 | 78.82 | 73.42 | 39.91 | |
| Mixup + CutMixBackbone=DeiT-Ti, Epochs=400, Batch size=128, Training Protocol=from scratch2026.06 | 78.75 | 72.93 | — | |
| TokenMixBackbone=DeiT-Ti, Epochs=400, Batch size=128, Training Protocol=from scratch2026.06 | 78.42 | 73.42 | — | |
| PuzzleMixBackbone=DeiT-Ti, Epochs=400, Batch size=128, Training Protocol=from scratch2026.06 | 78.32 | 72.89 | — | |
| Mixup + CutMixBackbone=DeiT-S, Epochs=400, Batch size=128, Training Protocol=from scratch2026.06 | 77.99 | 72.93 | 0.7 | |
| PuzzleMixBackbone=DeiT-S, Epochs=400, Batch size=128, Training Protocol=from scratch2026.06 | 77.27 | 72.89 | 273.28 | |
| FMixBackbone=DeiT-Ti, Epochs=400, Batch size=128, Training Protocol=from scratch2026.06 | 76.3 | 69.87 | — | |
| SnapMixBackbone=DeiT-Ti, Epochs=400, Batch size=128, Training Protocol=from scratch2026.06 | 76.17 | 71.13 | — | |
| SnapMixBackbone=DeiT-S, Epochs=400, Batch size=128, Training Protocol=from scratch2026.06 | 76.16 | 71.13 | 39.98 | |
| FMixBackbone=DeiT-S, Epochs=400, Batch size=128, Training Protocol=from scratch2026.06 | 73.76 | 69.87 | 6.8 | |
| MixupBackbone=DeiT-S, Epochs=400, Batch size=128, Training Protocol=from scratch2026.06 | 73.69 | 67.74 | 0.73 | |
| MixupBackbone=DeiT-Ti, Epochs=400, Batch size=128, Training Protocol=from scratch2026.06 | 72.58 | 67.74 | — | |
| ResizeMixBackbone=DeiT-Ti, Epochs=400, Batch size=128, Training Protocol=from scratch2026.06 | 71.85 | 67.33 | — | |
| ResizeMixBackbone=DeiT-S, Epochs=400, Batch size=128, Training Protocol=from scratch2026.06 | 70.8 | 67.33 | 1.05 | |
| VanillaBackbone=DeiT-S, Epochs=400, Batch size=128, Training Protocol=from scratch2026.06 | 66.91 | 61.71 | — | |
| VanillaBackbone=DeiT-Ti, Epochs=400, Batch size=128, Training Protocol=from scratch2026.06 | 66.46 | 61.71 | — |