VGG16 Implementaion in Keras

vgg16.py

def vgg16(input_shape=(224, 224, 3), classes=1000):
    '''
    ref: https://arxiv.org/abs/1409.1556
    default input_shape is 224, 224, 3
    default number of class is 1000
    '''
    # import dependencies
    # from keras.models import Model
    # import keras.layers as layer
    
    layer_input = layer.Input(shape=input_shape, name="input_layer")
    
    # block1
    # conv3-64
    # conv3-64
    # maxPool
    x = layer.Conv2D(filters=64, kernel_size=(3, 3), activation='relu', padding='same', name="block1_conv1")(layer_input)
    x = layer.Conv2D(filters=64, kernel_size=(3, 3), activation='relu', padding='same', name="block1_conv2")(x)
    x = layer.MaxPool2D(pool_size=(2, 2), strides=(2,2), name="block1_pool")(x)
    
    # block2
    # conv3-128
    # conv3-128
    # maxPool
    x = layer.Conv2D(filters=128, kernel_size=(3, 3), activation='relu', padding='same', name="block2_conv1")(x)
    x = layer.Conv2D(filters=128, kernel_size=(3, 3), activation='relu', padding='same', name="block2_conv2")(x)
    x = layer.MaxPool2D(pool_size=(2, 2), strides=(2,2), name="block2_pool")(x)
    
    # block3
    # conv3-256
    # conv3-256
    # maxPool
    x = layer.Conv2D(filters=256, kernel_size=(3, 3), activation='relu', padding='same', name="block3_conv1")(x)
    x = layer.Conv2D(filters=256, kernel_size=(3, 3), activation='relu', padding='same', name="block3_conv2")(x)
    x = layer.Conv2D(filters=256, kernel_size=(3, 3), activation='relu', padding='same', name="block3_conv3")(x)
    x = layer.MaxPool2D(pool_size=(2, 2), strides=(2,2), name="block3_pool")(x)
    
    # block4
    # conv3-512
    # conv3-512    
    # conv3-512    
    # maxPool
    x = layer.Conv2D(filters=512, kernel_size=(3, 3), activation='relu', padding='same', name="block4_conv1")(x)
    x = layer.Conv2D(filters=512, kernel_size=(3, 3), activation='relu', padding='same', name="block4_conv2")(x)
    x = layer.Conv2D(filters=512, kernel_size=(3, 3), activation='relu', padding='same', name="block4_conv3")(x)
    x = layer.MaxPool2D(pool_size=(2, 2), strides=(2,2), name="block4_pool")(x)
    
    # block5
    # conv3-512
    # conv3-512
    # conv3-512
    # maxPool
    x = layer.Conv2D(filters=512, kernel_size=(3, 3), activation='relu', padding='same', name="block5_conv1")(x)
    x = layer.Conv2D(filters=512, kernel_size=(3, 3), activation='relu', padding='same', name="block5_conv2")(x)
    x = layer.Conv2D(filters=512, kernel_size=(3, 3), activation='relu', padding='same', name="block5_conv3")(x)
    x = layer.MaxPool2D(pool_size=(2, 2), strides=(2,2), name="block5_pool")(x)

    
    # classification head
    x = layer.Flatten(name='flatten')(x)
    x = layer.Dense(units=4096, activation='relu', name="fc1")(x)
    x = layer.Dense(units=4096, activation='relu', name="fc2")(x)
    x = layer.Dense(units=classes, activation='softmax', name="fc3")(x)

    return Model(inputs=layer_input, outputs=x, name='vgg16')