input = Input(shape=(224,224,3), name='input_1')
    base_model = keras.applications.VGG16(weights='imagenet', include_top=False)
    for layer in base_model.layers:
        layer.trainable=False

    x = base_model(input)
    x = Conv2DTranspose(filters=256, kernel_size=(4,4), strides=(2, 2), activation='relu',
                        name='deconv1', padding='valid',
                        output_padding=(1,1),
                        kernel_initializer='random_uniform')(x)
    x = BatchNormalization()(x)
    x = Dropout(0.5)(x)
    x = Conv2D(filters=15, kernel_size=(1,1), name='final_conv', padding='same')(x)

<tf.Tensor 'vgg16/block5_pool/MaxPool:0' shape=(?, 7, 7, 512) dtype=float32>

Layer (type)                 Output Shape              Param #
=================================================================
input_1 (InputLayer)         (None, 224, 224, 3)       0
_________________________________________________________________
vgg16 (Model)                multiple                  14714688
_________________________________________________________________
deconv1 (Conv2DTranspose)    (None, 17, 17, 256)       2097408
_________________________________________________________________
batch_normalization_11 (Batc (None, 17, 17, 256)       1024
_________________________________________________________________
dropout_7 (Dropout)          (None, 17, 17, 256)       0
_________________________________________________________________
final_conv (Conv2D)          (None, 17, 17, 15)        3855
=================================================================
Total params: 16,816,975
Trainable params: 2,101,775
Non-trainable params: 14,715,200