Image Classification On Cifar 10 With Noisy
Metrics
Accuracy (under 20% Sym. label noise)
Accuracy (under 50% Sym. label noise)
Accuracy (under 80% Sym. label noise)
Accuracy (under 90% Sym. label noise)
Results
Performance results of various models on this benchmark
Model Name | Accuracy (under 20% Sym. label noise) | Accuracy (under 50% Sym. label noise) | Accuracy (under 80% Sym. label noise) | Accuracy (under 90% Sym. label noise) | Paper Title | Repository |
|---|---|---|---|---|---|---|
| SSR | 96.74% | 96.13% | 95.56% | 95.17% | SSR: An Efficient and Robust Framework for Learning with Unknown Label Noise | |
| PGDF (ResNet-18) | 96.7% | 96.3% | 94.7% | 84.0% | Sample Prior Guided Robust Model Learning to Suppress Noisy Labels | |
| C2D (DivideMix with SimCLR, ResNet-18) | 96.23 ± 0.09 | 95.15 ± 0.16 | 94.30 ± 0.12 | 93.42 ± 0.09 | Contrast to Divide: Self-Supervised Pre-Training for Learning with Noisy Labels | |
| C2D (ELR+ with SimCLR, ResNet-34) | 96.74 ± 0.12 | 95.55 ± 0.32 | 93.11 ± 0.70 | 89.30 ± 0.21 | Contrast to Divide: Self-Supervised Pre-Training for Learning with Noisy Labels |
0 of 4 row(s) selected.