def test_four_layers_wrong_paddig(self):
batch_size = 2
input_size = 256
images = tf.ones((batch_size, input_size, input_size, 3))
with self.assertRaises(TypeError):
discriminator.pix2pix_discriminator(
images, num_filters=[64, 128, 256, 512], padding=1.5)
def test_four_layers_negative_padding(self):
batch_size = 2
input_size = 256
images = tf.ones((batch_size, input_size, input_size, 3))
if tf.executing_eagerly():
exception_type = tf.errors.InvalidArgumentError
else:
exception_type = ValueError
with self.assertRaises(exception_type):
discriminator.pix2pix_discriminator(
images, num_filters=[64, 128, 256, 512], padding=-1)
def discriminator(image_batch, unused_conditioning=None):
"""A thin wrapper around the Pix2Pix discriminator to conform to TF-GAN."""
logits_4d, _ = d_module.pix2pix_discriminator(
image_batch, num_filters=[64, 128, 256, 512])
logits_4d.shape.assert_has_rank(4)
# Output of logits is 4D. Reshape to 2D, for TF-GAN.
logits_2d = tf.compat.v1.layers.flatten(logits_4d)
return logits_2d
def test_four_layers_no_padding(self):
batch_size = 2
input_size = 256
output_size = self._layer_output_size(input_size, pad=0)
output_size = self._layer_output_size(output_size, pad=0)
output_size = self._layer_output_size(output_size, pad=0)
output_size = self._layer_output_size(output_size, stride=1, pad=0)
output_size = self._layer_output_size(output_size, stride=1, pad=0)
images = tf.ones((batch_size, input_size, input_size, 3))
logits, end_points = discriminator.pix2pix_discriminator(
images, num_filters=[64, 128, 256, 512], padding=0)
self.assertListEqual([batch_size, output_size, output_size, 1],
logits.shape.as_list())
self.assertListEqual([batch_size, output_size, output_size, 1],
end_points['predictions'].shape.as_list())