Image Classification on CIFAR-10 semantic asymmetric noise (test)
89.96AccuracyScanMix
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| ScanMixBackbone=VGG, Noise ratio=34%2021.03 | 89.96 | — | — | — | — | — | — | — | |
| ScanMixBackbone=DenseNet, Noise ratio=32%2021.03 | 89.7 | — | — | — | — | — | — | — | |
| DivideMixBackbone=VGG, Noise ratio=34%2021.03 | 85.71 | — | — | — | — | — | — | — | |
| ScanMixBackbone=ResNet, Noise ratio=38%2021.03 | 85.58 | — | — | — | — | — | — | — | |
| DivideMixBackbone=DenseNet, Noise ratio=32%2021.03 | 84.57 | — | — | — | — | — | — | — | |
| DivideMixBackbone=ResNet, Noise ratio=38%2021.03 | 81.61 | — | — | — | — | — | — | — | |
| Backward + ROGBackbone=VGG, Noise ratio=34%2021.03 | 70.18 | — | — | — | — | — | — | — | |
| Bootstrap + ROGBackbone=VGG, Noise ratio=34%2021.03 | 70.11 | — | — | — | — | — | — | — | |
| Forward + RoGBackbone=VGG, Noise ratio=34%2021.03 | 70.09 | — | — | — | — | — | — | — | |
| CE + ROGBackbone=VGG, Noise ratio=34%2021.03 | 70.04 | — | — | — | — | — | — | — | |
| Backward + ROGBackbone=DenseNet, Noise ratio=32%2021.03 | 68.66 | — | — | — | — | — | — | — | |
| D2L + ROGBackbone=DenseNet, Noise ratio=32%2021.03 | 68.57 | — | — | — | — | — | — | — | |
| Bootstrap + ROGBackbone=DenseNet, Noise ratio=32%2021.03 | 68.38 | — | — | — | — | — | — | — | |
| CE + ROGBackbone=DenseNet, Noise ratio=32%2021.03 | 68.33 | — | — | — | — | — | — | — | |
| Forward + RoGBackbone=DenseNet, Noise ratio=32%2021.03 | 68.2 | — | — | — | — | — | — | — | |
| Forward + RoGBackbone=ResNet, Noise ratio=38%2021.03 | 64.24 | — | — | — | — | — | — | — | |
| CE + ROGBackbone=ResNet, Noise ratio=38%2021.03 | 64.15 | — | — | — | — | — | — | — | |
| Bootstrap + ROGBackbone=ResNet, Noise ratio=38%2021.03 | 64.03 | — | — | — | — | — | — | — | |
| Backward + ROGBackbone=ResNet, Noise ratio=38%2021.03 | 63.45 | — | — | — | — | — | — | — | |
| D2L + ROGBackbone=ResNet, Noise ratio=38%2021.03 | 60.25 | — | — | — | — | — | — | — | |
| D2L + ROGBackbone=VGG, Noise ratio=34%2021.03 | 59.94 | — | — | — | — | — | — | — | |
| Co-teaching2021.05 | — | 84.5 | 70.69 | 54.29 | 48.76 | 46.4 | 45.66 | 44.39 | |
| Cross Entropy2021.05 | — | 88.3 | 84.28 | 77.4 | 67.54 | 61.72 | 57.24 | 52.7 | |
| DivideMix2021.05 | — | 93.95 | 84.43 | 73.73 | 60.13 | 53.18 | 50.6 | 49.42 | |
| JoCoR2021.05 | — | 76.57 | 67.74 | 56.54 | 48.52 | 46.22 | 43.74 | 43.06 | |
| Joint Optimization2021.05 | — | 92.01 | 89.56 | 84.56 | 78.21 | 76.7 | 76.44 | 76 | |
| Loss Correction2021.05 | — | 90.87 | 87.21 | 74.63 | 58.82 | 58.06 | 53.72 | 53.04 | |
| Mixup2021.05 | — | 90.53 | 86.59 | 78.67 | 69.19 | 62.52 | 57.9 | 53.3 | |
| PUDistill2021.05 | — | 90 | 88.22 | 85.98 | 84.41 | 83.77 | 83 | 82.63 |