Table 2.

CNN layers and hyperparameters

Layer	Hyperparameters
Conv2D	32 filters, 3 × 3 filter size, ReLU activation, same padding, followed by batch normalization
MaxPool2D	3 × 3 pool size to reduce image spatial dimensions quickly from 96 × 96 to 32 × 32
Dropout (Core Layer)	0.25 Neurons
Conv2D	64 filters, 3 × 3 filter size, ReLU activation, same padding
Conv2D	64 filters, 3 × 3 filter size, ReLU activation, following the same padding, batch normalization is performed
MaxPool2D	2 × 2 pool size
Dropout (Core Layer)	0.25 Neurons
Conv2D	128 filters, 3 × 3 filter, ReLU activation, following the same padding, batch normalization is performed
Conv2D	128 filters, 3 × 3 filter size, ReLU activation, same padding followed by batch normalization
MaxPool2D	2 × 2 pool size
Dropout (Core Layer)	0.25 Neurons
Flatten (Core Layer)	–
Dense	1024 Units, ReLU sctivation, and batch normalization
Dropout (Core Layer)	0.5 Neurons
Dense	7 Units, softmax activation