Robust Image Classification on CIFAR-10 (test) (Symmetric & Asymmetric Noise)
96.75Accuracy (Symmetric Noise, η=0.2)Ours (Modified Model)
Evaluation results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Ours (Modified Model)Model Variant=Modified, Test-Time Augmentation=true2022.10 | 96.75 | — | — | — | 94.98 | — | — | 96.42 | 96.16 | 93.93 | |
| ContrastToDivide2022.10 | 96.4 | — | — | — | 94.4 | — | — | 93.5 | 95.3 | 93.6 | |
| AugDesc2022.10 | 96.3 | — | — | — | 93.8 | — | — | 94.6 | 95.4 | 91.9 | |
| PropMix2022.10 | 96.09 | — | — | — | 93.77 | — | — | 94.64 | 95.53 | 93.2 | |
| Ours (Modified Model)Model Variant=Modified2022.10 | 95.99 | — | — | — | 94.48 | — | — | 95.85 | 95.59 | 93.52 | |
| ELR2022.10 | 95.8 | — | — | — | 93.3 | — | — | 93 | 94.8 | 78.7 | |
| DivideMix2022.10 | 95.7 | — | — | — | 92.9 | — | — | 92.1 | 94.4 | 75.4 | |
| Ours (Normal Model)Model Variant=Normal, Test-Time Augmentation=true2022.10 | 95.47 | — | — | — | 94.9 | — | — | 95.14 | 95.39 | 92.32 | |
| Ours (Normal Model)Model Variant=Normal2022.10 | 95.04 | — | — | — | 94.51 | — | — | 94.91 | 95.13 | 91.72 | |
| FasTENBackbone=ResNet-342021.11 | 91.94 | — | 90.07 | 86.78 | 79.52 | 92.29 | — | 90.43 | — | — | |
| MSLCBackbone=ResNet-342021.11 | 90.94 | — | 88.36 | 83.93 | 64.9 | 91.45 | — | 89.26 | — | — | |
| MLOCBackbone=ResNet-342021.11 | 90.5 | — | 87.2 | 81.95 | 54.64 | 91.15 | — | 89.35 | — | — | |
| Deep kNNBackbone=ResNet-342021.11 | 90.02 | — | 87.27 | 82.8 | 68.3 | 89.97 | — | 84.56 | — | — | |
| MW-NetBackbone=ResNet-342021.11 | 89.76 | — | 86.52 | 81.68 | 56.56 | 91.31 | — | 88.69 | — | — | |
| MLaCBackbone=ResNet-342021.11 | 89.75 | — | 86.63 | 82.2 | 71.94 | 91.45 | — | 90.26 | — | — | |
| GLCBackbone=ResNet-342021.11 | 89.66 | — | 85.3 | 80.34 | 67.44 | 91.56 | — | 89.76 | — | — | |
| L2RWBackbone=ResNet-342021.11 | 88.26 | — | 83.76 | 74.54 | 42.6 | 88.79 | — | 85.86 | — | — | |
| SL2019.08 | 87.63 | 89.28 | 85.34 | 80.07 | 53.81 | 88.24 | 85.36 | 80.64 | — | — | |
| Forward2019.08 | 85.83 | 89.39 | 81.37 | 73.59 | 47.1 | 87.68 | 84.86 | 85.73 | — | — | |
| GCE2019.08 | 84.86 | 86.76 | 82.42 | 75.2 | 40.81 | 84.61 | 82.11 | 75.32 | — | — | |
| Bootstrap2019.08 | 83.95 | 88.77 | 79.97 | 71.65 | 41.44 | 86.57 | 86.86 | 79.76 | — | — | |
| LSR2019.08 | 83.49 | 88.57 | 78.41 | 67.38 | 36.3 | 85.38 | 84.86 | 77.88 | — | — | |
| CE2019.08 | 82.96 | 89.26 | 78.7 | 66.62 | 34.8 | 85.98 | 82.89 | 78.51 | — | — | |
| Cross-Entropy2022.10 | 82.7 | — | — | — | 26.1 | — | — | 72.3 | 57.9 | 16.8 | |
| D2L2019.08 | 81.13 | 86.66 | 76.8 | 60.67 | 19.83 | 82.72 | 80.41 | 73.33 | — | — |