Fig 1.
Technical flow chart of damage detection.
Table 1.
Network structure table of the classifier.
Fig 2.
The network structure of SegFormer [45].
Fig 3.
Sample dataset.
Table 2.
Composition of the dataset.
Fig 4.
The training curves of classification model. (a) Concrete spalling classifier. (b) Crack classifier.
Fig 5.
Example of classification results of the original image.
Fig 6.
The training curves of segmentation model.
(a) Concrete spalling segmentation. (b) Crack segmentation.
Fig 7.
Example of segmentation results.
Fig 8.
The skeleton line of crack binary image.
Fig 9.
Crack size information.
Fig 10.
Confusion matrix of safety evaluation.
Table 3.
Hyperparameter for each model.
Table 4.
Comparative experiment.
Fig 11.
Confusion matrix of several comparison models.
(a) GB. (b) SVM. (c) MLP. (d) DT.