Image Classification on CIFAR-100 (Top-1 Accuracy Curve Evaluation)
70.72Top-1 Acc (p=0.0)Fine-tuning
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Fine-tuningBackbone=ResNet-32, Noise Type=Flip2022.11 | 70.72 | 56.98 | 46.37 | — | — | |
| MW-netBackbone=ResNet-32, Noise Type=Flip, Implementation=Our implementation2022.11 | 70.57 | 64.13 | 51.23 | 19.89 | 7.42 | |
| CrossEntropyBackbone=ResNet-32, Noise Type=Flip, Implementation=Reported from original paper2022.11 | 70.5 | 50.86 | 43.01 | — | — | |
| MentorNetBackbone=ResNet-32, Noise Type=Flip2022.11 | 70.24 | 61.97 | 52.66 | — | — | |
| CrossEntropyBackbone=ResNet-32, Noise Type=Flip, Implementation=Our implementation2022.11 | 70.18 | 65.02 | 50.25 | 18.67 | 4.32 | |
| MW-netBackbone=ResNet-32, Noise Type=Flip, Implementation=Reported from original paper2022.11 | 70.11 | 64.22 | 58.64 | — | — | |
| Focal LossBackbone=ResNet-32, Noise Type=Flip2022.11 | 70.02 | 61.87 | 54.13 | — | — | |
| Reed-HardBackbone=ResNet-32, Noise Type=Flip2022.11 | 69.02 | 60.27 | 50.4 | — | — | |
| Advisor Network (Ours)Backbone=ResNet-32, Noise Type=Flip, Implementation=Ours2022.11 | 68.93 | 63.54 | 59.07 | 56.13 | 20.29 | |
| D2LBackbone=ResNet-32, Noise Type=Flip2022.11 | 68.11 | 63.48 | 51.83 | — | — | |
| Self-pacedBackbone=ResNet-32, Noise Type=Flip2022.11 | 67.55 | 63.63 | 53.51 | — | — | |
| GLCBackbone=ResNet-32, Noise Type=Flip2022.11 | 65.42 | 63.07 | 62.22 | — | — | |
| L2RWBackbone=ResNet-32, Noise Type=Flip2022.11 | 64.11 | 57.47 | 50.98 | — | — | |
| Co-teachingBackbone=ResNet-32, Noise Type=Flip2022.11 | 63.31 | 54.13 | 44.85 | — | — | |
| S-ModelBackbone=ResNet-32, Noise Type=Flip2022.11 | 51.46 | 45.45 | 43.8 | — | — |