from keras.layers import Input
from keras.layers import Conv3D
from keras.layers import ReLU
from keras.layers import Dropout
from keras.layers import GaussianNoise
from keras.layers import MaxPooling3D
from keras.regularizers import l2

inflow = Input(shape=input_shape, name='input')

args = dict(
    use_bias=False,
    kernel_regularizer=l2(penalty),
    kernel_initializer='he_uniform'
)

x = Conv3D(24, 7, name='conv0', **args)(inflow)
x = LayerNormalization(name='norm0')(x)
x = ReLU(name='nonlin0')(x)
x = MaxPooling3D(padding='same', name='pool0')(x)
x = Dropout(dropout, name='drop0')(x)

x = Conv3D(32, 5, name='conv1', **args)(x)
x = LayerNormalization(name='norm1')(x)
x = ReLU(name='nonlin1')(x)
x = MaxPooling3D(padding='same', name='pool1')(x)
x = Dropout(dropout, name='drop1')(x)

x = Conv3D(48, 3, name='conv2', **args)(x)
x = LayerNormalization(name='norm2')(x)
x = ReLU(name='nonlin2')(x)
x = MaxPooling3D(padding='same', name='pool2')(x)
x = Dropout(dropout, name='drop2')(x)

x = Conv3D(64, 3, name='conv3', **args)(x)
x = LayerNormalization(name='norm3')(x)
x = ReLU(name='nonlin3')(x)
x = MaxPooling3D(padding='same', name='pool3')(x)
x = Dropout(dropout, name='drop3')(x)

x = Conv3D(96, 2, name='conv4', **args)(x)
x = LayerNormalization(name='norm4')(x)
x = ReLU(name='nonlin4')(x)

outflow = GaussianNoise(stddev=noise, name='noise4')(x)

model = keras.models.Model(inputs=inflow, outputs=outflow, name=name)