Image Classification on CIFAR-10 instance-dependent noise (IDN) (test)
90.31Accuracy (η=0.2)T-Revision
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| T-RevisionBackbone=ResNet-50, Pre-trained=ImageNet2021.02 | 90.31 | 84.99 | 72.06 | |
| HOC LocalBackbone=ResNet-50, Pre-trained=ImageNet2021.02 | 90.03 | 85.49 | 77.4 | |
| Co-teaching+Backbone=ResNet-50, Pre-trained=ImageNet2021.02 | 89.82 | 73.44 | 63.61 | |
| HOC GlobalBackbone=ResNet-50, Pre-trained=ImageNet2021.02 | 89.71 | 84.62 | 70.67 | |
| Peer LossBackbone=ResNet-50, Pre-trained=ImageNet2021.02 | 89.52 | 83.44 | 75.15 | |
| CEε+MAE2025.08 | 89.27 | 85.26 | 74.32 | |
| NCE+RCE2025.08 | 89.06 | 85.07 | 70.45 | |
| LDR-KL2025.08 | 88.99 | 84.1 | 63.11 | |
| NCE+AGCE2025.08 | 88.9 | 85.16 | 72.68 | |
| Co-teachingBackbone=ResNet-50, Pre-trained=ImageNet2021.02 | 88.84 | 72.61 | 63.76 | |
| JoCoRBackbone=ResNet-50, Pre-trained=ImageNet2021.02 | 88.82 | 71.13 | 63.88 | |
| LDMIBackbone=ResNet-50, Pre-trained=ImageNet2021.02 | 88.67 | 83.65 | 69.82 | |
| ForwardBackbone=ResNet-50, Pre-trained=ImageNet2021.02 | 87.87 | 79.81 | 68.32 | |
| GCE2025.08 | 86.95 | 79.35 | 52.3 | |
| SCE2025.08 | 86.79 | 74.56 | 49.63 | |
| CE (Standard)Backbone=ResNet-50, Pre-trained=ImageNet2021.02 | 85.66 | 76.89 | 60.29 | |
| LqBackbone=ResNet-50, Pre-trained=ImageNet2021.02 | 85.66 | 75.24 | 61.3 | |
| CE2025.08 | 75.05 | 57.27 | 37.62 |