Algorithm 4. DSCNet Model with Self-Attention

1: Default configurations: 2: num layers ← 4 4: kernel size ← 3 5: activation function ← ‘relu’ 6: dropoutrate ← 0.5 7: Initialize the model 8:   model ← initialize model() 9: for i = 1 to num layers do 10:  model.add layer (Conv2D(num filters[i], kernel size, activation = activation     function)) 11:  model.add layer (MaxPooling2D(pool size = 2)) 12:  model.add layer (Dropout(dropout rate)) 13:  model.add_layer(SelfAttention()) 14: end for 15: model.add layer (Flatten()) 16: model.add layer (Dense(128, activation = activation function)) 17: model.add layer (Dropout(dropout rate)) 18: model.add layer (Dense(1, activation = ‘linear’))