Image Classification on STL-10 (Top-5 Accuracy)
99.72Top-5 AccuracyOurs
Evaluation results
| Method | Links | |
|---|---|---|
| OursDistortion Type=Random missing pixels (RM), Evaluation Protocol=Linear probing, Pre-trained on=ImageNet-1002021.10 | 99.72 | |
| OursDistortion Type=Additive Gaussian noise (GN), Evaluation Protocol=Linear probing, Pre-trained on=ImageNet-1002021.10 | 99.48 | |
| BaselineDistortion Type=Additive Gaussian noise (GN), Evaluation Protocol=Linear probing, Pre-trained on=ImageNet-1002021.10 | 99.41 | |
| BaselineDistortion Type=Random missing pixels (RM), Evaluation Protocol=Linear probing, Pre-trained on=ImageNet-1002021.10 | 99.38 |