dcgan.py

import tensorflow as tf
from layers import inception_module, downsample, conv, initializer

SIZE=128

def encoder(x):
    with tf.variable_scope("encoder"):
        net = inception_module(x, 32)
        net = downsample(net, 2)
        net = inception_module(net, 32)
        net = downsample(net, 2)
        net = inception_module(net, 16)
        net = downsample(net, 2)
        net = conv(net, 32, 1)
        net = tf.reshape(net, [-1, net.shape[1]*net.shape[2]*net.shape[3]])
        net = tf.layers.dense(net, units=(SIZE//16*SIZE//16*32), activation=tf.nn.sigmoid, kernel_initializer=initializer)
        return net

def decoder(x):
    with tf.variable_scope("decoder"):
        net = tf.reshape(x, shape=[-1, SIZE//16,SIZE//16,32])
        net = tf.layers.conv2d_transpose(net, filters=512, kernel_size=5, strides=2, padding="same", kernel_initializer=initializer)
        net = tf.layers.batch_normalization(net)
        net = tf.nn.relu(net)

        net = tf.layers.conv2d_transpose(net, filters=256, kernel_size=5, strides=2, padding="same", kernel_initializer=initializer)
        net = tf.layers.batch_normalization(net)
        net = tf.nn.relu(net)

        net = tf.layers.conv2d_transpose(net, filters=64 , kernel_size=5, strides=2, padding="same", kernel_initializer=initializer)
        net = tf.layers.batch_normalization(net)
        net = tf.nn.relu(net)

        net = tf.layers.conv2d_transpose(net, filters=32 , kernel_size=5, strides=2, padding="same", kernel_initializer=initializer)
        net = tf.layers.batch_normalization(net)
        net = tf.nn.relu(net)

        net = tf.layers.conv2d_transpose(net, filters=3, kernel_size=5, activation=tf.nn.sigmoid, padding="same")
        return net