Image Classification on Controlled Noisy Web Labels (CNWL) Red Mini-ImageNet 1.0 (test)
51.42Accuracy (20% Noise)FaMUS
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| FaMUSBackbone=PreAct ResNet-182021.04 | 51.42 | 48.06 | 45.1 | 35.5 | 45.02 | |
| MentorMixBackbone=PreAct ResNet-182021.04 | 51.02 | 47.14 | 43.8 | 33.46 | 43.85 | |
| DivideMixBackbone=PreAct ResNet-182021.04 | 50.96 | 46.72 | 43.14 | 34.5 | 43.83 | |
| MixupBackbone=PreAct ResNet-182021.04 | 49.1 | 46.4 | 40.58 | 33.58 | 42.41 | |
| Cross-entropyBackbone=PreAct ResNet-182021.04 | 47.36 | 42.7 | 37.3 | 29.76 | 39.28 |