Image Classification on CIFAR100 Corrupted (test)
59.67AccuracyTEA
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| TEABackbone=ResNet-50, Normalization=GroupNorm2023.11 | 59.67 | 0.8932 | — | |
| SHOTBackbone=ResNet-50, Normalization=GroupNorm2023.11 | 58.31 | 0.9345 | — | |
| EATABackbone=ResNet-50, Normalization=GroupNorm2023.11 | 56.76 | 0.9628 | — | |
| ETABackbone=ResNet-50, Normalization=GroupNorm2023.11 | 56.72 | 0.9637 | — | |
| PLBackbone=ResNet-50, Normalization=GroupNorm2023.11 | 56.68 | 0.9602 | — | |
| TENTBackbone=ResNet-50, Normalization=GroupNorm2023.11 | 56.34 | 0.9688 | — | |
| SARBackbone=ResNet-50, Normalization=GroupNorm2023.11 | 55.28 | 0.9933 | — | |
| SourceBackbone=ResNet-50, Normalization=GroupNorm2023.11 | 54.98 | 1 | — | |
| Deep EnsemblesPass Type=Multi-Pass, Backbone=ResNet-28-10, Number of Seeds=102021.10 | 54.1 | 0.138 | 2.281 | |
| Geometric Sensitivity DecompositionPass Type=Multi-Pass, Backbone=ResNet-28-10, Number of Seeds=10, Variant=Ensemble, Tuning=Grid-Searched2021.10 | 54.1 | 0.086 | 2.042 | |
| VanillaPass Type=Single-Pass, Backbone=ResNet-28-10, Number of Seeds=102021.10 | 50.5 | 0.239 | 2.756 | |
| DUQPass Type=Single-Pass, Backbone=ResNet-28-10, Number of Seeds=102021.10 | 50.4 | 0.281 | 2.841 | |
| Geometric Sensitivity DecompositionPass Type=Single-Pass, Backbone=ResNet-28-10, Number of Seeds=10, Tuning=Grid-Searched2021.10 | 49.8 | 0.081 | 2.23 | |
| Geometric Sensitivity DecompositionPass Type=Single-Pass, Backbone=ResNet-28-10, Number of Seeds=10, Tuning=Optimized2021.10 | 49.8 | 0.088 | 2.236 | |
| SNGPPass Type=Single-Pass, Backbone=ResNet-28-10, Number of Seeds=102021.10 | 49 | 0.117 | 2.626 | |
| MC DropoutPass Type=Multi-Pass, Backbone=ResNet-28-10, Number of Seeds=102021.10 | 42.6 | 0.202 | 2.881 |