Number of parameters in batch normalization and the axis as 3

Question

Here is the code that i found from Andrew Ng deeplearning specialization course:

def HappyModel(input_shape):

X_input = Input(input_shape)

X = ZeroPad