def __init__(self, filters, sn=False, **kwargs): super(uk_5s1_2D, self).__init__() self.filters_zp = filters self.l1_conv2d = BaseLayers.Conv2DTranspose(filters=filters, kernel_size=[5, 5], strides=[1, 1], padding="SAME", use_bias=False, sn=sn, **kwargs) self.l2_norm = BaseLayers.BatchNormalization() self.l3_activation = BaseLayers.LeakyReLU()
def __init__(self,filters,sn=False,activation=None,**kwargs): super(lats_up,self).__init__() self.filters_zp = filters self.l1_conv2d = BaseLayers.Conv2DTranspose(filters=filters,kernel_size=[5,5],strides=[2,2],padding="SAME",use_bias=False,sn=sn,**kwargs) if activation == None: self.l2_activation = tf.keras.layers.Activation("linear") elif activation.lower()=="relu": self.l2_activation = BaseLayers.ReLU() elif activation.lower()=="leaky_relu": self.l2_activation = BaseLayers.LeakyReLU() elif activation.lower()=="sigmoid": self.l2_activation = tf.keras.layers.Activation("sigmoid") elif activation.lower()=="tanh": self.l2_activation = tf.keras.layers.Activation("tanh") else: raise ValueError("Un supported activation"+activation)