Image Classification on CIFAR-10 (test) (Noise Robustness Accuracy)
93.2Accuracy (Sym. Noise Rate 0.2)SPR
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| SPRBackbone=ResNet-182022.03 | 93.2 | 91 | 82.7 | 64.1 | 92.8 | 91.3 | 89 | |
| TopoFilterBackbone=ResNet-182022.03 | 90.2 | 87.2 | 80.5 | 45.7 | 90.5 | 89.7 | 87.9 | |
| Co-teaching+Backbone=ResNet-182022.03 | 89.8 | 86.1 | 74 | 17.9 | 89.4 | 87.1 | 71.3 | |
| Co-teachingBackbone=ResNet-182022.03 | 89.2 | 86.4 | 79 | 22.9 | 90 | 88.2 | 78.4 | |
| ROGBackbone=ResNet-182022.03 | 89.2 | 83.5 | 77.9 | 29.1 | 89.6 | 88.4 | 86.2 | |
| SCEBackbone=ResNet-182022.03 | 89.2 | 85.3 | 78 | 44.4 | 88.7 | 86.3 | 81.4 | |
| GCEBackbone=ResNet-182022.03 | 88.7 | 84.7 | 76.1 | 41.7 | 88.1 | 86 | 81.4 | |
| PENCILBackbone=ResNet-182022.03 | 88.2 | 86.6 | 74.3 | 45.3 | 90.2 | 88.3 | 84.5 | |
| MentorNetBackbone=ResNet-182022.03 | 88.1 | 81.4 | 70.4 | 31.3 | 86.3 | 84.8 | 78.7 | |
| IterNLDBackbone=ResNet-182022.03 | 87.9 | 83.7 | 74.1 | 38 | 89.3 | 88.8 | 85 | |
| DecouplingBackbone=ResNet-182022.03 | 87.4 | 83.3 | 73.8 | 36 | 89.3 | 88.1 | 85.1 | |
| BootstrapBackbone=ResNet-182022.03 | 86.4 | 82.5 | 75.2 | 42.1 | 88.8 | 87.5 | 85.1 | |
| ForgettingBackbone=ResNet-182022.03 | 86 | 82.1 | 75.5 | 41.3 | 89.5 | 88.2 | 85 | |
| StandardBackbone=ResNet-182022.03 | 85.7 | 81.8 | 73.7 | 42 | 88 | 86.4 | 84.9 | |
| ForwardBackbone=ResNet-182022.03 | 85.7 | 81 | 73.3 | 31.6 | 88.5 | 87.3 | 85.3 |