Convolutional Neural Network with Multiscale Fusion and Attention Mechanism for Skin Diseases Assisted Diagnosis

Table 1

We compare the results of different rate combinations and introduce scSE into the multiscale feature before and after fusion. It is used to compare the impact of single and composite feature maps on global accuracy.

Methods (scSE)	Dice	IoU
MSFA-Net (I and II)	0.9066	0.8311
MSFA-Net (I and II)	0.9129	0.9112
MSFA-Net (I and II)	0.9248	0.8852
MSFA-Net (I)	0.9020	0.8374
MSFA-Net (I)	0.9026	0.8228
MSFA-Net (I)	0.9075	0.8653
MSFA-Net (II)	0.8520	0.7972
MSFA-Net (II)	0.9199	0.8573
MSFA-Net (II)	0.9271	0.9128