Table 4 Comparison results of model performance across different architectures under augmentation.

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

Metrics

ViT

Swin Transformer

PVT

Raw

Aug

Gain

Raw

Aug

Gain

Raw

Aug

Gain

 

AP50

83.07

86.43

3.36

84.28

87.25

2.97

83.4

86.45

3.05

 

AP50:95

58.53

63.81

5.28

59.78

64.52

4.74

58.92

62.87

3.95

 

IOU

76.22

81.41

5.19

77.15

82.56

5.41

76.58

80.74

4.16

 

\(\:{AP}_{S}\)

52.21

59.35

7.14

53.4

60.94

7.54

53.15

58.76

5.61

 

\(\:{AP}_{M}\)

80.13

83.42

3.29

80.41

84.93

4.52

80.22

83.33

3.11

 

\(\:{AP}_{L}\)

87.02

90.12

3.1

88.23

91.59

3.36

87.47

89.91

2.44

 
Back to article page