def test_random_height_inference(self):
with CustomObjectScope({'RandomHeight': image_preprocessing.RandomHeight}):
input_images = np.random.random((2, 5, 8, 3)).astype(np.float32)
expected_output = input_images
with tf_test_util.use_gpu():
layer = image_preprocessing.RandomHeight(.5)
actual_output = layer(input_images, training=0)
self.assertAllClose(expected_output, actual_output)
def test_valid_random_height(self):
# need (maxval - minval) * rnd + minval = 0.6
mock_factor = 0
with test.mock.patch.object(
gen_stateful_random_ops, 'stateful_uniform', return_value=mock_factor):
with tf_test_util.use_gpu():
img = np.random.random((12, 5, 8, 3))
layer = image_preprocessing.RandomHeight(.4)
img_out = layer(img, training=True)
self.assertEqual(img_out.shape[1], 3)
def _run_test(self, factor):
np.random.seed(1337)
num_samples = 2
orig_height = 5
orig_width = 8
channels = 3
with tf_test_util.use_gpu():
img = np.random.random((num_samples, orig_height, orig_width, channels))
layer = image_preprocessing.RandomHeight(factor)
img_out = layer(img, training=True)
self.assertEqual(img_out.shape[0], 2)
self.assertEqual(img_out.shape[2], 8)
self.assertEqual(img_out.shape[3], 3)
def test_random_height_shorter_numeric(self):
for dtype in (np.int64, np.float32):
with tf_test_util.use_gpu():
input_image = np.reshape(np.arange(0, 8), (4, 2, 1)).astype(dtype)
layer = image_preprocessing.RandomHeight(
factor=(-.5, -.5), interpolation='nearest')
output_image = layer(np.expand_dims(input_image, axis=0))
# pyformat: disable
expected_output = np.asarray([
[2, 3],
[6, 7]
]).astype(dtype)
# pyformat: enable
expected_output = np.reshape(expected_output, (1, 2, 2, 1))
self.assertAllEqual(expected_output, output_image)
def test_random_height_longer_numeric(self):
for dtype in (np.int64, np.float32):
with tf_test_util.use_gpu():
input_image = np.reshape(np.arange(0, 6), (2, 3, 1)).astype(dtype)
layer = image_preprocessing.RandomHeight(factor=(1., 1.))
output_image = layer(np.expand_dims(input_image, axis=0))
# pyformat: disable
expected_output = np.asarray([
[0, 1, 2],
[0.75, 1.75, 2.75],
[2.25, 3.25, 4.25],
[3, 4, 5]
]).astype(dtype)
# pyformat: enable
expected_output = np.reshape(expected_output, (1, 4, 3, 1))
self.assertAllEqual(expected_output, output_image)
def test_random_height_longer_numeric(self):
for dtype in (np.int64, np.float32):
with tf_test_util.use_gpu():
input_image = np.reshape(np.arange(0, 6), (2, 3, 1)).astype(dtype)
layer = image_preprocessing.RandomHeight(factor=(1., 1.))
# Return type of RandomHeight() is float32 if `interpolation` is not
# set to `ResizeMethod.NEAREST_NEIGHBOR`; cast `layer` to desired dtype.
output_image = math_ops.cast(layer(np.expand_dims(input_image, axis=0)),
dtype=dtype)
# pyformat: disable
expected_output = np.asarray([
[0, 1, 2],
[0.75, 1.75, 2.75],
[2.25, 3.25, 4.25],
[3, 4, 5]
]).astype(dtype)
# pyformat: enable
expected_output = np.reshape(expected_output, (1, 4, 3, 1))
self.assertAllEqual(expected_output, output_image)
def test_config_with_custom_name(self): layer = image_preprocessing.RandomHeight(.5, name='image_preproc') config = layer.get_config() layer_1 = image_preprocessing.RandomHeight.from_config(config) self.assertEqual(layer_1.name, layer.name)
def test_random_height_invalid_factor(self):
with self.assertRaises(ValueError):
image_preprocessing.RandomHeight((-1.5, .4))
