Table 1 Comparative evaluation metrics of different models on the BioHit dataset and the self-built dataset.

From: An improved lightweight tongue segmentation model with self-attention parallel network and progressive upsampling

Network

BioHit

Self-Built Dataset

MPA (%)

MIou (%)

Dice (%)

MPA (%)

MIou (%)

Dice (%)

FCN

91.75

94.92

92.32

90.30

95.02

93.86

UNet

84.33

92.76

86.55

94.94

97.33

96.80

PSPNet

93.96

97.35

94.95

95.10

96.78

97.00

DeepLabv3

95.59

98.29

96.44

95.83

97.93

97.41

Segformer

95.14

97.82

95.97

96.68

98.31

97.92

PAPU_TonSeg

97.56

98.60

97.99

97.32

98.64

98.32

  1. Significant values are in bold.
Back to article page