Table 4 Results of underwater pipelines recognition using the proposed method in HOMOMO, Roboflow, and YouTube including their standard deviations.

From: Integrating simplified Swin-T with modified EFS-Net for attention-guided underwater pipelines segmentation in complex underwater environments

Criteria

YouTube

Roboflow

HOMOMO

Mean IoU

81.32% ± 6.17

85.4% ± 4.21

98.44%  ± 0.91

Mean accuracy

93.0% ± 4.21

94.6% ± 2.50

99.5% ± 0.62

Mean precision

85.73% ± 5.31

86.98% ± 3.81

98.98% ± 0.88

Mean recall

84.76% ± 5.78

86.76% ± 3.92

98.52% ± 0.78

Mean specificity

85.41% ± 3.98

87.36% ± 2.1

99.1% ± 0.69

Mean F1-score

85.24% ± 5.43

86.86% ± 3.86

98.74% ± 0.84

Mean F-boundary

70.01% ± 8.35

75.99% ± 7.81

82.01% ± 1.3

Back to article page