Image Classification on ImageNet-9 Mixed-Rand v1 (test)
58.2AccuracyCF + F Combination
Evaluation results
| Method | Links | |
|---|---|---|
| CF + F CombinationCF variant=CAGAN, F variant=Shuffle2021.06 | 58.2 | |
| CF + F + Saliency CombinationCF variant=CAGAN, F variant=Shuffle, Saliency regularization=true2021.06 | 58 | |
| Factual Augmentation (F)Variant=Shuffle2021.06 | 57.1 | |
| F + Saliency CombinationF variant=Shuffle, Saliency regularization=true2021.06 | 57.1 | |
| Factual Augmentation (F)Variant=Random2021.06 | 56.4 | |
| Counterfactual Augmentation (CF)Variant=CAGAN2021.06 | 53.3 | |
| CF + Saliency CombinationCF variant=CAGAN, Saliency regularization=true2021.06 | 53.2 | |
| Label Smoothing (LS)epsilon=0.12021.06 | 52.5 | |
| Saliency regularization (Sal)lambda=10^-42021.06 | 52.4 | |
| Mixupalpha=0.22021.06 | 52.2 | |
| Counterfactual Augmentation (CF)Variant=Random2021.06 | 52.1 | |
| Counterfactual Augmentation (CF)Variant=Shuffle2021.06 | 51.8 | |
| Factual Augmentation (F)Variant=FGSM, epsilon=0.52021.06 | 51.6 | |
| Original Baseline2021.06 | 51.5 | |
| Counterfactual Augmentation (CF)Variant=Grey2021.06 | 51.4 | |
| Counterfactual Augmentation (CF)Variant=Tile2021.06 | 46.5 |