Image Classification on TinyImageNet (test) (Adversarial Error Bounds)
0.966Adversarial Error Bound Start (L^e) (x10^-7)Normal
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| NormalModel Architecture=ConvNet2024.03 | 0.966 | 0.633 | 28.71 | 0 | |
| NormalModel Architecture=ResNet182024.03 | 1.131 | 0.528 | 28.34 | 0 | |
| SEPposModel Architecture=ResNet182024.03 | 1.967 | 1.435 | 25.96 | 10.83 | |
| SEPposModel Architecture=ConvNet2024.03 | 2.242 | 1.571 | 24.83 | 9.63 | |
| MATModel Architecture=ConvNet2024.03 | 2.426 | 1.728 | 25.13 | 9.55 | |
| MATModel Architecture=ResNet182024.03 | 3.119 | 2.349 | 26.34 | 10.81 | |
| SEPnegModel Architecture=ConvNet2024.03 | 3.873 | 2.61 | 24.31 | 9.61 | |
| SEPnegModel Architecture=ResNet182024.03 | 4.413 | 3.016 | 26.11 | 10.74 |