from keras.layers import Conv2D, Activation, MaxPooling2D, Flatten, Dense
from keras.layers import Input, ZeroPadding2D
from keras.layers import Dropout, BatchNormalization
from keras.models import Model
from keras.utils import plot_model


input_shape=(64, 64, 3)

# Define the input placeholder as a tensor with shape input_shape.
X_input = Input(input_shape)

# Zero-Padding: pads the border of X_input with zeroes
X = ZeroPadding2D((3, 3))(X_input)

# CONV -> BN -> RELU
X = Conv2D(filters=32, kernel_size=(7, 7), strides=(1, 1), padding='same', name='conv0')(X)
X = BatchNormalization(axis=3, name='bn0')(X)
X = Activation('relu')(X)

# MAXPOOL
# padding: valid (default) or same
X = MaxPooling2D(pool_size=(2, 2), strides=(2,2), padding='valid', name='max_pool1')(X)

# CONV -> RELU
X = Conv2D(filters=64, kernel_size=(3, 3), strides=(1, 1), padding='same', activation='relu', name='conv1')(X)

# MAXPOOL
# strides default to pool_size
X = MaxPooling2D(pool_size=(2, 2), name='max_pool2')(X)

# Flatten X: convert X to a vector with shape (batch_size, h * w * c)
X = Flatten()(X)

# Fully connected layer
X = Dense(units=512, activation='relu', name='fc0')(X)
X = Dropout(0.25)(X)
# softmax -> multi-class classification, units = num_classes
# sigmoid -> binary classification, units = 1
X = Dense(units=1, activation='sigmoid', name='fc1')(X)

# Create model
model = Model(inputs=X_input, outputs=X, name='Model')

model.summary()

plot_model(model, show_shapes=True, show_layer_names=True, to_file='model.png')