Medical Image Classification on DermaMNIST 40% Noise
98.7SensitivityUNICON+CS
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| UNICON+CSCS Loss=True2026.04 | 98.7 | 47.4 | 73.1 | 70.6 | 47.6 | |
| DivideMixCS Loss=False2026.04 | 98 | 47.1 | 72.5 | 69 | 47.1 | |
| UNICONCS Loss=False2026.04 | 98 | 48.5 | 73.2 | 70.1 | 47.8 | |
| Co-teaching+CSCS Loss=True2026.04 | 93.4 | 46.4 | 69.9 | 71 | 45.1 | |
| BaselineCS Loss=False2026.04 | 84.9 | 43.6 | 64.3 | 68.8 | 40.7 | |
| GMM FilterCS Loss=False2026.04 | 76 | 59.1 | 67.6 | 71.7 | 44.2 | |
| GMM Filter+CSCS Loss=True2026.04 | 73 | 35.9 | 54.4 | 58.3 | 33.4 | |
| DivideMix+CSCS Loss=True2026.04 | 66.3 | 59.4 | 62.9 | 65.7 | 39.8 | |
| Baseline+CSCS Loss=True2026.04 | 64 | 46.9 | 55.4 | 58.6 | 33.5 | |
| Co-teachingCS Loss=False2026.04 | 50.8 | 79.3 | 65 | 73.4 | 43 |