Skip to main content

Table 12 The accuracies (%) of models for each category on the test sets of the four datasets

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

Dataset

Model

B

M

E

N

L

LDWBC

ResNet [45]

80.95

83.17

97.89

98.51

96.22

DenseNet [46]

90.48

81.73

98.95

99.21

95.88

Inception v3 [54]

42.86

37.50

52.63

93.48

91.46

Jiang [55]

90.48

74.04

100

99.44

97.60

Sharma [56]

64.29

50.48

87.37

99.21

97.84

Our model

95.24

90.38

100

99.44

96.86

LISC

ResNet [45]

100

62.50

100

100

100

DenseNet [46]

100

87.50

100

100

100

Inception v3 [54]

91.67

50.00

63.64

85.71

81.82

Jiang [55]

100

87.50

100

100

100

Sharma [56]

100

75.00

100

100

100

Our model

100

87.50

100

100

100

BCCD

ResNet [45]

–

75.00

79.55

84.38

100

DenseNet [46]

–

75.00

84.42

89.21

100

Inception v3 [54]

–

51.95

65.58

69.24

64.38

Jiang [55]

–

72.73

84.58

89.86

100

Sharma [56]

–

73.86

85.23

89.21

99.84

Our model

–

74.84

85.23

93.72

100

Raabin

ResNet [45]

100

88.03

98.14

95.83

98.74

DenseNet [46]

100

92.31

97.83

96.88

98.36

Inception v3 [54]

92.13

81.62

86.96

90.00

90.81

Jiang [55]

100

92.74

98.76

95.08

98.45

Sharma [56]

100

85.04

96.27

95.94

98.16

Our model

100

94.87

99.07

98.65

99.52

  1. Best results are in bold; B basophil, M monocyte, E eosinophil, N neutrophil, L lymphocyte