Image Classification on CIFAR-10H (test)
89.68AccuracyNonlinear transformation framework
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Nonlinear transformation frameworkBackbone=VGG19, Pre-trained=True (ImageNet)2024.06 | 89.68 | — | — | — | 58 | 65 | |
| OPABackbone=VGG19, Pre-trained=True (ImageNet)2024.06 | 89.1 | — | — | — | — | — | |
| Mixupseeds=102026.05 | 88.24 | 0.5526 | 9.77 | 0.2499 | — | — | |
| SLSseeds=102026.05 | 86.95 | 0.5052 | 1.86 | 0.3946 | — | — | |
| Soft labelsseeds=102026.05 | 86.87 | 0.5096 | 1.85 | 0.3909 | — | — | |
| Label smoothingseeds=102026.05 | 85.9 | 0.6263 | 5.98 | 0.2117 | — | — | |
| Majority voteseeds=102026.05 | 85.7 | 0.7284 | 7.04 | 0.2902 | — | — | |
| Adversarial trainingBackbone=VGG19, Pre-trained=True (ImageNet), Defense Method=PGD2024.06 | 80.41 | — | — | — | — | — | |
| No defenseBackbone=VGG19, Pre-trained=True (ImageNet)2024.06 | 19 | — | — | — | — | — |