Table 9 Edge AI model performance comparison on prototype application.
From: MTAKD: multi-teacher agreement knowledge distillation for edge AI skin disease diagnosis
Model | ISIC 2019 | Fitzpatrick17k-C | ||
|---|---|---|---|---|
Model size (KB) | Inference time (ms) | Model size (KB) | Inference time (ms) | |
RegNetY32GF | 559,310 | 37,054 ± 375 | 559,523 | 37,118 ± 561 |
DenseNet201 | 74,659 | 1,424 ± 5 | 74,871 | 1,429 ± 3 |
Xception | 85,336 | 1,663 ± 26 | 85,549 | 1,656 ± 24 |
InceptionV3 | 89,201 | 846 ± 7 | 89,413 | 846 ± 4 |
NASNetMobile | 18,792 | 223 ± 3 | 19,004 | 224 ± 3 |
EfficientNetV2B0 | 25,431 | 305 ± 4 | 25,644 | 305 ± 1 |
MTAKD Student Model | 11,230 | 105 ± 3 | 11,442 | 105 ± 4 |
