Table 2 Results of different segmentation methods on the nuclear cladding coating surface crack dataset.

From: A crack detection model fusing local details and global context for nuclear cladding coating surfaces

Method

MIoU(%)

P(%)

R(%)

F1-Score(%)

U-Net

82.84

81.91

79.21

80.54

DeepLabv3+

80.65

72.75

82.84

77.47

SETR

78.95

74.62

75.63

75.12

SegFormer

80.52

79.25

75.87

77.52

Mask2Former

82.22

83.82

84.60

84.20

CrackFormer-II

87.19

91.09

80.75

90.92

SAM

83.23

87.79

85.32

86.53

CrackCTFuse

92.70

88.47

97.01

92.54

Back to article page