def FeatureExtractor(self, X, reuse = False):
input_X = utils.NormalizeImage(X)
with tf.variable_scope('feature_extractor_conv1',reuse = reuse):
h_conv1 = layers.conv2d(input_X, self.ef_dim, 3, stride=1,
activation_fn=tf.nn.relu, weights_initializer=self.initializer)
h_conv1 = layers.conv2d(h_conv1, self.ef_dim, 3, stride=1,
activation_fn=tf.nn.relu, weights_initializer=self.initializer)
h_conv1 = layers.max_pool2d(h_conv1, [2, 2], 2, padding='SAME')
with tf.variable_scope('feature_extractor_conv2',reuse = reuse):
h_conv2 = layers.conv2d(h_conv1, self.ef_dim * 2, 3, stride=1,
activation_fn=tf.nn.relu, weights_initializer=self.initializer)
h_conv2 = layers.conv2d(h_conv2, self.ef_dim * 2, 3, stride=1,
activation_fn=tf.nn.relu, weights_initializer=self.initializer)
h_conv2 = layers.max_pool2d(h_conv2, [2, 2], 2, padding='SAME')
with tf.variable_scope('feature_extractor_conv3',reuse = reuse):
h_conv3 = layers.conv2d(h_conv2, self.ef_dim * 4, 3, stride=1,
activation_fn=tf.nn.relu, weights_initializer=self.initializer)
h_conv3 = layers.conv2d(h_conv3, self.ef_dim * 4, 3, stride=1,
activation_fn=tf.nn.relu, weights_initializer=self.initializer)
h_conv3 = layers.max_pool2d(h_conv3, [2, 2], 2, padding='SAME')
with tf.variable_scope('feature_extractor_fc1', reuse = reuse):
fc_input = layers.flatten(h_conv3)
fc_1 = layers.fully_connected(inputs=fc_input, num_outputs=self.latent_dim,
activation_fn=None, weights_initializer=self.initializer)
features = fc_1
return features
def create_operators(self):
size = 224
img_mean = [0.485, 0.456, 0.406]
img_std = [0.229, 0.224, 0.225]
img_scale = 1.0 / 255.0
decode_op = utils.DecodeImage()
resize_op = utils.ResizeImage(resize_short=256)
crop_op = utils.CropImage(size=(size, size))
normalize_op = utils.NormalizeImage(scale=img_scale,
mean=img_mean,
std=img_std)
totensor_op = utils.ToTensor()
return [decode_op, resize_op, crop_op, normalize_op, totensor_op]
def create_operators(interpolation=1):
size = 224
img_mean = [0.485, 0.456, 0.406]
img_std = [0.229, 0.224, 0.225]
img_scale = 1.0 / 255.0
resize_op = utils.ResizeImage(resize_short=256,
interpolation=interpolation)
crop_op = utils.CropImage(size=(size, size))
normalize_op = utils.NormalizeImage(scale=img_scale,
mean=img_mean,
std=img_std)
totensor_op = utils.ToTensor()
return [resize_op, crop_op, normalize_op, totensor_op]