Table 3 Quantitative assessment of the machine annotation stevia construction of a confusion matrix.

From: Benthic and coral reef community field data for Heron Reef, Southern Great Barrier Reef, Australia, 2002–2018

Camera

SONY

Canon

Lumix

Olympus

Years

2002–2006

2007–2010

2011–2016

2017–2018

Overall Accuracy (%)

79.1

81.8

73.9

79.8

User’s Accuracy (%)

Hard Coral

79.9

83.6

83.2

88.2

Rock

77.2

79.3

71.2

74.4

Rubble

68.0

68.8

61.5

25.0

Sand

85.7

90.3

87.2

93.9

Algae

85.7

79.4

74.4

71.4

Other

52.4

33.3

57.3

61.7

Producer’s Accuracy (%)

Hard Coral

76.0

72.7

72.5

70.2

Rock

89.2

92.6

90.5

94.8

Rubble

5.3

15.6

4.7

10.2

Sand

92.1

94.5

89.8

91.8

Algae

6.8

42.8

19.4

24.2

Other

23.7

18.7

24.0

33.5

# Points

8,000

7,150

18,500

3,300

  1. For each camera used, machine annotation (modelled data) of 2.5% of all the photoquadrats captured was compared with manual annotation (reference data) of the same validation data set in a using standard confusion matrix3. From this, the overall accuracy and individual class accuracies were calculated following a well-documented approach3.