Learning with noisy labels on Food-101N noise ratio ~20% (test)
88Top-1 Test AccuracySURE
Evaluation Results
| Method | Links | |
|---|---|---|
| SUREBackbone=ResNet-50, Extra self-supervised loss=false2024.03 | 88 | |
| Jigsaw-ViTBackbone=DeiT-S, Extra self-supervised loss=true2024.03 | 86.7 | |
| WarPIBackbone=ResNet-50, Extra self-supervised loss=false2024.03 | 85.9 | |
| PLCBackbone=ResNet-50, Extra self-supervised loss=false2024.03 | 85.3 | |
| NRankBackbone=ResNet-50, Extra self-supervised loss=false2024.03 | 85.2 | |
| SMPBackbone=ResNet-50, Extra self-supervised loss=false2024.03 | 85.1 | |
| MWNetBackbone=ResNet-50, Extra self-supervised loss=false2024.03 | 84.7 | |
| CleanNetBackbone=ResNet-50, Extra self-supervised loss=false2024.03 | 83.5 | |
| CEBackbone=ResNet-50, Extra self-supervised loss=false2024.03 | 81.7 |