Table 9 Results of the best model, grouped by different image conditions at AP50 level.

From: Automated non-PPE detection on construction sites using YOLOv10 and transformer architectures for surveillance and body worn cameras with benchmark datasets

Image conditions

Ranges

Class

Non-helmet

Non-mask

Non-vest

Non-glove

Non-shoes

mAP

Illumination

High

92.91

89.49

88.27

86.44

84

88.22

Low

91.7

86.6

86.41

84.64

82.75

86.42

Visual range

Near-field view

92.69

89.63

88.5

86.18

84.15

88.23

Far-field view

91.85

86.97

86.19

84.6

82.46

86.41

Space

Indoor

93.27

88.8

87.95

86.33

84.42

88.15

Outdoor

92.44

86.86

86.17

84.75

82.21

86.49

Overlap of worker

Non-overlap

93.16

88.72

87.62

86.8

84.82

88.22

Overlap

92.36

86.86

85.65

84.73

82.49

86.42

Density of worker

Scatter

92.78

89.15

87.69

86.86

84.21

88.13

Grouped

92.32

86.97

86.18

84.7

82.34

86.51

Number of workers

Single

93.45

89.26

88.42

86.68

84.38

88.44

Mutiple

92.41

86.57

85.68

84.03

82.31

86.2

Back to article page