Image Classification on TinyImageNet (test) (Corruption Metrics)
74.69AccuracySYN+NST+TA (Ours)
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| SYN+NST+TA (Ours)Backbone=WRN-28-4, Training Epochs=6002025.12 | 74.69 | 49.77 | 50.15 | 45.93 | |
| Synthetic2025.12 | 73.37 | — | 33.43 | 39.88 | |
| SYN+NST+TA2025.12 | 73.07 | — | 49.26 | 44.72 | |
| SYN+NST+TA + Random Erasing + Noise + Feature Noise2025.12 | 72.43 | — | 49.24 | 44.68 | |
| SYN+NST+TA + Random Erasing + Noise2025.12 | 71.8 | — | 48.36 | 43.34 | |
| Noisy Feature Mixup + TrivialAugment2025.12 | 69.93 | — | 40 | 40.43 | |
| TrivialAugment + Random Erasing + Noise2025.12 | 69.23 | — | 40.33 | 38.89 | |
| Mixup + Cutmix2025.12 | 68.26 | — | 30.32 | 38.17 | |
| TrivialAugment2025.12 | 68.01 | — | 39.64 | 37.41 | |
| NoisyMixTraining Strategy=300 epoch rerun without early stopping2025.12 | 67.83 | — | 40.85 | — | |
| NoisyMixBackbone=WRN-28-4, Training Epochs=6002025.12 | 67.82 | 42.13 | 41.31 | — | |
| Random Erasing2025.12 | 64.56 | — | 26.38 | 32.09 | |
| Random Erasing + Noise2025.12 | 63.88 | — | 27.54 | 33.26 | |
| AugMix2025.12 | 62.36 | — | 36.13 | 33.98 | |
| AugMix + DeepAugment2025.12 | 62.07 | — | 36.29 | 38.06 | |
| Baseline2025.12 | 61.95 | — | 24.67 | 31.03 |