Application of U-Net with Global Convolution Network Module in Computer-Aided Tongue Diagnosis

Table 1

Model comparison with or without the GCN module and BR module.

Model	PA (pixel accuracy)	Loss (%)	MIoU (mean intersection over union)
SEResNet_Unet	98.5	4.96	38.0
SEResNet_GCN_Unet	98.5	5.10	43.1
VGGNet_Unet	98.6	4.32	42.0
VGGNet_GCN_Unet	98.6	4.84	46.2
GoogLeNet_Unet	98.6	4.11	45.5
GoogLeNet_GCN_Unet	98.7	4.48	47.3