Medical Image Classification on PathMNIST 40% Noise
96.8SensitivityCo-teaching+CS
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Co-teaching+CSCS Loss=True2026.04 | 96.8 | 19.2 | 58 | 75.9 | 41.5 | |
| UNICON+CSCS Loss=True2026.04 | 90.4 | 62.6 | 75.7 | 81.2 | 59.9 | |
| Baseline+CSCS Loss=True2026.04 | 85.5 | 22.4 | 54 | 64.9 | 38.5 | |
| GMM Filter+CSCS Loss=True2026.04 | 85.5 | 76.5 | 81 | 86.9 | 64.8 | |
| BaselineCS Loss=False2026.04 | 81.4 | 92.2 | 86.8 | 92.2 | 78.5 | |
| GMM FilterCS Loss=False2026.04 | 77.6 | 95.9 | 86.8 | 94.6 | 81.2 | |
| Co-teachingCS Loss=False2026.04 | 77.4 | 96.6 | 87 | 96.3 | 82 | |
| DivideMixCS Loss=False2026.04 | 72.8 | 95.7 | 84.2 | 92.3 | 77.7 | |
| UNICONCS Loss=False2026.04 | 72.8 | 95.7 | 84.2 | 92.3 | 77.7 | |
| DivideMix+CSCS Loss=True2026.04 | 71.5 | 79.9 | 75.7 | 79.8 | 59.9 |