Identification of Civil Infrastructure Damage Using Ensemble Transfer Learning Model

Table 3

Structure of custom CNN model.

Layer	Layer type	No. of nodes	Kernel size	Activation	Dropout rate
0	Input shape	64 × 64	—	—	—
1	Conv1	16	3 × 3	ReLU	—
2	Conv2	16	3 × 3	ReLU	—
3	Max-pool	—	2 × 2	—	0.2
4	Conv3	32	3 × 3	ReLU	—
5	Conv4	32	3 × 3	ReLU	—
6	Max-pool	—	2 × 2	—	0.2
7	Conv5	64	3 × 3	ReLU	—
8	Conv6	64	3 × 3	ReLU	—
9	Max-pool	—	2 × 2	—	—
10	FC	512	—	ReLU	—
11	Output	5	—	Softmax	—