Skip to main content

Table 11 The comparing results (%) of different methods on the LISC, BCCD, and Raabin test sets

From: Accurate classification of white blood cells by coupling pre-trained ResNet and DenseNet with SCAM mechanism

Dataset

Model

OA

AP

AR

AF1

LISC

ResNet [45]

93.88

95.48

92.50

92.98

DenseNet [46]

97.96

98.33

97.50

97.80

Inception v3 [54]

75.51

74.88

74.57

73.74

Jiang [55]

97.96

98.46

97.50

97.87

Sharma [56]

95.92

96.79

95.00

95.47

Our model

97.96

98.33

97.50

97.80

BCCD

ResNet [45]

84.71

87.06

84.73

85.15

DenseNet [46]

87.14

89.36

87.16

87.48

Inception v3 [54]

62.80

67.71

62.79

63.52

Jiang [55]

86.77

89.28

86.79

87.10

Sharma [56]

87.02

89.15

87.03

87.31

Our model

88.44

90.84

88.45

88.73

Raabin

ResNet [45]

96.36

92.87

96.15

94.28

DenseNet [46]

97.12

94.02

97.07

95.42

Inception v3 [54]

89.56

78.39

88.31

82.47

Jiang [55]

96.13

91.69

97.00

93.97

Sharma [56]

95.99

92.62

95.08

93.50

Our model

98.71

97.18

98.42

97.78

  1. Best results are in bold