Skip to main content

Table 3 Patch-level Validation ACA (%) of CNN Models Trained on The Original/Refined Training Sets

From: Reverse active learning based atrous DenseNet for pathological image classification

 

BACH

CCG

UCSB

 

Original

Refined

Original

Refined

Original

Refined

AlexNet [1]

86.28

90.77

75.67

82.24

76.51

93.78

VGG-16 [10]

90.83

91.79

84.63

90.02

83.53

97.44

ResNet-50 [12]

89.65

92.17

79.88

82.31

78.74

96.82

ResNet-101 [12]

89.05

91.17

80.06

83.47

77.82

96.78

DenseNet [13]

90.39

93.29

77.87

84.41

78.93

96.79

ADN (ours)

91.93

94.10

85.48

92.05

85.69

97.63

  1. Best accuracy is in Bold.