Identification of Civil Infrastructure Damage Using Ensemble Transfer Learning Model

Table 4

Performance analysis proposed method with the different models.

Ref	Classification types	Methods	Accuracy	Precision	Recall
[32]	Pothole classification	ResNet18	90.5	87.2	80.2
[33]	Pothole classification	Image processing	73.5	80.0	73.3
[34]	Pothole classification	LS-SVM	85.2	72.7	76.3
[34]	Pothole classification	ANN	88.7	85.7	78.2
[23]	Road damage detection	CNN	81.4	82.9	81.1
Present study	Bridge and road damage (5 classes)	Custom CNN	86.0	78.0	94.0
Present study	Bridge and road damage (5 classes)	Xception	84.0	79.0	89.0
Present study	Bridge and road damage (5 classes)	AlexNet	86.0	78.0	94.0
Present study	Bridge and road damage (5 classes)	Ensemble model	87.1	84.92	83.53