Skip to main content

Table 3 Performance of the proposed stomata detection algorithm

From: A generalised approach for high-throughput instance segmentation of stomata in microscope images

Dataset

Quality

Known to model

Num. of stomata

Precision (%)

Recall (%)

F-Score (%)

Gymnosperm 400×

Med–High

Yes

944

95.87

98.41

97.12

Gymnosperm 100×: low

Low

Yes

10597

98.89

91.92

95.28

Gymnosperm 100×: high

High

Yes

7713

98.15

94.30

96.18

Poplar

High

Yes

5042

98.34

96.11

97.22

Cuticle: low

Low

Partially

8181

93.46

73.51

82.29

Cuticle: med

Medium

Partially

2631

94.80

89.43

92.04

Ginkgo

High

Partially

2802

96.02

82.65

88.84

USNM/USBG: low

Low

Partially

2569

92.70

70.65

80.19

USNM/USBG: med

Medium

Partially

16083

95.20

82.31

88.30

Betula nana

Low–Med

No

683

85.62

75.25

80.06

Eucalyptus

Medium

No

1088

93.22

83.46

88.07

Ferns: low

Low

No

964

78.91

51.24

62.14

Ferns: med

Medium

No

713

90.15

74.47

81.56

Grass

Low–Med

No

3288

85.20

55.66

67.32

UNSW-2019

Med–High

No

2242

91.53

85.77

88.56

Google Images

Medium

No

1496

97.52

76.34

85.64