Image Classification on CIFAR-100 Symmetric Noise (test)
81.9AccuracySANM(C2D)
Evaluation Results
| Method | Links | |
|---|---|---|
| SANM(C2D)Noise ratio=20%, Noise type=Symmetric, Base model=C2D2023.02 | 81.9 | |
| SANM(DivideMix)Noise ratio=20%, Noise type=Symmetric, Base model=DivideMix2023.02 | 81.2 | |
| AugDescNoise ratio=20%, Noise type=Symmetric2023.02 | 79.5 | |
| SANM(C2D)Noise ratio=50%, Noise type=Symmetric, Base model=C2D2023.02 | 79.3 | |
| C2DNoise ratio=20%, Noise type=Symmetric2023.02 | 78.6 | |
| SANM(DivideMix)Noise ratio=50%, Noise type=Symmetric, Base model=DivideMix2023.02 | 78.2 | |
| DivideMixNoise ratio=20%, Noise type=Symmetric2023.02 | 77.3 | |
| AugDescNoise ratio=50%, Noise type=Symmetric2023.02 | 77.2 | |
| Sel-CL+Noise ratio=20%, Noise type=Symmetric2023.02 | 76.5 | |
| C2DNoise ratio=50%, Noise type=Symmetric2023.02 | 76.4 | |
| MOIT+Noise ratio=20%, Noise type=Symmetric2023.02 | 75.9 | |
| DivideMixNoise ratio=50%, Noise type=Symmetric2023.02 | 74.6 | |
| M-correctionNoise ratio=20%, Noise type=Symmetric2023.02 | 73.9 | |
| Sel-CL+Noise ratio=50%, Noise type=Symmetric2023.02 | 72.4 | |
| SANM(C2D)Noise ratio=80%, Noise type=Symmetric, Base model=C2D2023.02 | 71.6 | |
| MOIT+Noise ratio=50%, Noise type=Symmetric2023.02 | 70.6 | |
| PENCILNoise ratio=20%, Noise type=Symmetric2023.02 | 69.4 | |
| SANM(DivideMix)Noise ratio=80%, Noise type=Symmetric, Base model=DivideMix2023.02 | 68.7 | |
| Meta-LearningNoise ratio=20%, Noise type=Symmetric2023.02 | 68.5 | |
| GCENoise ratio=20%, Noise type=Symmetric2023.02 | 68.1 | |
| ELRBackbone=ResNet-34, Noise Rate (epsilon)=0.42021.06 | 68.03 | |
| MixupNoise ratio=20%, Noise type=Symmetric2023.02 | 67.8 | |
| C2DNoise ratio=80%, Noise type=Symmetric2023.02 | 67.7 | |
| AugDescNoise ratio=80%, Noise type=Symmetric2023.02 | 66.4 | |
| M-correctionNoise ratio=50%, Noise type=Symmetric2023.02 | 66.1 | |
| Co-teaching+Noise ratio=20%, Noise type=Symmetric2023.02 | 65.6 | |
| Peer LossBackbone=ResNet-34, Noise Rate (epsilon)=0.42021.06 | 62.16 | |
| Cross-Entropy (CE)Noise ratio=20%, Noise type=Symmetric2023.02 | 62 | |
| SANM(C2D)Noise ratio=90%, Noise type=Symmetric, Base model=C2D2023.02 | 61.9 | |
| ELRBackbone=ResNet-34, Noise Rate (epsilon)=0.62021.06 | 60.49 | |
| DivideMixNoise ratio=80%, Noise type=Symmetric2023.02 | 60.2 | |
| Sel-CL+Noise ratio=80%, Noise type=Symmetric2023.02 | 59.6 | |
| AUMBackbone=ResNet-34, Noise Rate (epsilon)=0.42021.06 | 59.29 | |
| Meta-LearningNoise ratio=50%, Noise type=Symmetric2023.02 | 59.2 | |
| C2DNoise ratio=90%, Noise type=Symmetric2023.02 | 58.7 | |
| Negative/Not Label Smoothing (NLS)Backbone=ResNet-34, Noise Rate (epsilon)=0.42021.06 | 58.47 | |
| PENCILNoise ratio=50%, Noise type=Symmetric2023.02 | 57.5 | |
| MixupNoise ratio=50%, Noise type=Symmetric2023.02 | 57.3 | |
| Label Smoothing (LS)Backbone=ResNet-34, Noise Rate (epsilon)=0.42021.06 | 55.17 | |
| Peer LossBackbone=ResNet-34, Noise Rate (epsilon)=0.62021.06 | 53.72 | |
| GCENoise ratio=50%, Noise type=Symmetric2023.02 | 53.3 | |
| FLCBackbone=ResNet-34, Noise Rate (epsilon)=0.42021.06 | 53.04 | |
| Co-teaching+Noise ratio=50%, Noise type=Symmetric2023.02 | 51.8 | |
| APLBackbone=ResNet-34, Noise Rate (epsilon)=0.42021.06 | 51.63 | |
| SCEBackbone=ResNet-34, Noise Rate (epsilon)=0.42021.06 | 49.34 | |
| Sel-CL+Noise ratio=90%, Noise type=Symmetric2023.02 | 48.8 | |
| Cross EntropyBackbone=ResNet-34, Noise Rate (epsilon)=0.42021.06 | 48.2 | |
| M-correctionNoise ratio=80%, Noise type=Symmetric2023.02 | 48.2 | |
| MOIT+Noise ratio=80%, Noise type=Symmetric2023.02 | 47.6 | |
| BootstrapBackbone=ResNet-34, Noise Rate (epsilon)=0.42021.06 | 47.28 | |
| Cross-Entropy (CE)Noise ratio=50%, Noise type=Symmetric2023.02 | 46.7 | |
| Negative/Not Label Smoothing (NLS)Backbone=ResNet-34, Noise Rate (epsilon)=0.62021.06 | 46.58 | |
| AUMBackbone=ResNet-34, Noise Rate (epsilon)=0.62021.06 | 44.05 | |
| SANM(DivideMix)Noise ratio=90%, Noise type=Symmetric, Base model=DivideMix2023.02 | 43.5 | |
| Meta-LearningNoise ratio=80%, Noise type=Symmetric2023.02 | 42.4 | |
| APLBackbone=ResNet-34, Noise Rate (epsilon)=0.62021.06 | 42.31 | |
| MOIT+Noise ratio=90%, Noise type=Symmetric2023.02 | 41.8 | |
| Label Smoothing (LS)Backbone=ResNet-34, Noise Rate (epsilon)=0.62021.06 | 41.63 | |
| FLCBackbone=ResNet-34, Noise Rate (epsilon)=0.62021.06 | 41.59 | |
| AugDescNoise ratio=90%, Noise type=Symmetric2023.02 | 41.2 | |
| SCEBackbone=ResNet-34, Noise Rate (epsilon)=0.62021.06 | 38.87 | |
| Cross EntropyBackbone=ResNet-34, Noise Rate (epsilon)=0.62021.06 | 38.27 | |
| BootstrapBackbone=ResNet-34, Noise Rate (epsilon)=0.62021.06 | 35.81 | |
| DivideMixNoise ratio=90%, Noise type=Symmetric2023.02 | 31.5 | |
| PENCILNoise ratio=80%, Noise type=Symmetric2023.02 | 31.1 | |
| MixupNoise ratio=80%, Noise type=Symmetric2023.02 | 30.8 | |
| Co-teaching+Noise ratio=80%, Noise type=Symmetric2023.02 | 27.9 | |
| M-correctionNoise ratio=90%, Noise type=Symmetric2023.02 | 24.3 | |
| GCENoise ratio=80%, Noise type=Symmetric2023.02 | 22.1 | |
| Cross-Entropy (CE)Noise ratio=80%, Noise type=Symmetric2023.02 | 19.9 | |
| Meta-LearningNoise ratio=90%, Noise type=Symmetric2023.02 | 19.5 | |
| PENCILNoise ratio=90%, Noise type=Symmetric2023.02 | 15.3 | |
| MixupNoise ratio=90%, Noise type=Symmetric2023.02 | 14.6 | |
| Co-teaching+Noise ratio=90%, Noise type=Symmetric2023.02 | 13.7 | |
| Cross-Entropy (CE)Noise ratio=90%, Noise type=Symmetric2023.02 | 10.1 | |
| GCENoise ratio=90%, Noise type=Symmetric2023.02 | 8.9 |