Intelligent Defect Identification Based on PECT Signals and an Optimized Two-Dimensional Deep Convolutional Network

Table 4

Identification accuracy and computational cost for different defect identification pipelines.

TFR	GoogLeNet		Inception V3		ResNet50		AlexNet		VGG16
	Accuracy (%)	Time (s)	Accuracy (%)	Accuracy (%)	Accuracy (%)	Accuracy (%)	Accuracy (%)	Time (s)	Accuracy (%)	Time (s)
STFT	74.12 ± 0.85	34.4	74.36 ± 1.67	74.36 ± 1.67	74.36 ± 1.67	74.36 ± 1.67	75.50 ± 1.57	14.5	85.16 ± 0.80	191.8
PSWVD	73.78 ± 0.89	35.6	71.42 ± 1.21	71.42 ± 1.21	71.42 ± 1.21	71.42 ± 1.21	89.92 ± 1.44	15.4	90.03 ± 1.80	193.4
EEMD	77.90 ± 1.00	36.7	73.52 ± 1.32	73.52 ± 1.32	73.52 ± 1.32	73.52 ± 1.32	80.42 ± 1.26	16.7	96.57 ± 0.48	196.7
SST	72.77 ± 0.52	37.4	74.37 ± 1.89	74.37 ± 1.89	74.37 ± 1.89	74.37 ± 1.89	80.33 ± 0.45	17.5	97.88 ± 0.35	208.5
GWT	78.16 ± 1.03	36.2	74.79 ± 1.04	74.79 ± 1.04	74.79 ± 1.04	74.79 ± 1.04	91.10 ± 0.31	14.1	92.46 ± 0.43	189.6