Table 1 Classification performance on the ISIC 2019 dataset.
From: MTAKD: multi-teacher agreement knowledge distillation for edge AI skin disease diagnosis
Model | Number of parameters | Accuracy (%) | Precision (%) | Recall (%) | F1-score (%) |
|---|---|---|---|---|---|
RegNetY32GF | 143,397,218 | 86.48 ± 0.41 | 86.33 ± 0.36 | 86.48 ± 0.41 | 86.32 ± 0.39 |
DenseNet201 | 19,309,640 | 85.37 ± 0.48 | 85.41 ± 0.46 | 85.37 ± 0.48 | 85.31 ± 0.47 |
Xception | 21,914,672 | 81.33 ± 0.54 | 81.05 ± 0.65 | 81.33 ± 0.54 | 81.02 ± 0.62 |
InceptionV3 | 22,855,976 | 82.45 ± 0.38 | 82.41 ± 0.42 | 82.45 ± 0.38 | 82.29 ± 0.38 |
NASNetMobile | 4,815,004 | 80.53 ± 0.63 | 80.14 ± 0.81 | 80.53 ± 0.63 | 80.08 ± 0.80 |
EfficientNetV2B0 | 6,579,288 | 80.57 ± 0.60 | 80.33 ± 0.54 | 80.57 ± 0.60 | 79.98 ± 0.64 |
MobileNetV2 | 2,917,960 | 80.17 ± 0.72 | 80.00 ± 0.76 | 80.23 ± 0.82 | 79.86 ± 0.78 |
