def testInt32Axis(self):
x = np.array([[1, 0, 0], [1, 0, 0], [2, 0, 0]])
with self.test_session() as sess:
y0, idx0 = gen_array_ops.unique_v2(x, axis=[0])
tf_y0, tf_idx0 = sess.run([y0, idx0])
y1, idx1 = gen_array_ops.unique_v2(x, axis=[1])
tf_y1, tf_idx1 = sess.run([y1, idx1])
self.assertAllEqual(tf_y0, np.array([[1, 0, 0], [2, 0, 0]]))
self.assertAllEqual(tf_idx0, np.array([0, 0, 1]))
self.assertAllEqual(tf_y1, np.array([[1, 0], [1, 0], [2, 0]]))
self.assertAllEqual(tf_idx1, np.array([0, 1, 1]))
def testInt32Axis(self):
x = np.array([[1, 0, 0], [1, 0, 0], [2, 0, 0]])
with self.test_session() as sess:
y0, idx0 = gen_array_ops.unique_v2(x, axis=[0])
tf_y0, tf_idx0 = sess.run([y0, idx0])
y1, idx1 = gen_array_ops.unique_v2(x, axis=[1])
tf_y1, tf_idx1 = sess.run([y1, idx1])
self.assertAllEqual(tf_y0, np.array([[1, 0, 0], [2, 0, 0]]))
self.assertAllEqual(tf_idx0, np.array([0, 0, 1]))
self.assertAllEqual(tf_y1, np.array([[1, 0], [1, 0], [2, 0]]))
self.assertAllEqual(tf_idx1, np.array([0, 1, 1]))
def testInt32Axis(self):
for dtype in [np.int32, np.int64]:
x = np.array([[1, 0, 0], [1, 0, 0], [2, 0, 0]])
with self.cached_session() as sess:
y0, idx0 = gen_array_ops.unique_v2(x, axis=np.array([0], dtype))
tf_y0, tf_idx0 = self.evaluate([y0, idx0])
y1, idx1 = gen_array_ops.unique_v2(x, axis=np.array([1], dtype))
tf_y1, tf_idx1 = self.evaluate([y1, idx1])
self.assertAllEqual(tf_y0, np.array([[1, 0, 0], [2, 0, 0]]))
self.assertAllEqual(tf_idx0, np.array([0, 0, 1]))
self.assertAllEqual(tf_y1, np.array([[1, 0], [1, 0], [2, 0]]))
self.assertAllEqual(tf_idx1, np.array([0, 1, 1]))
def testInt32Axis(self):
for dtype in [np.int32, np.int64]:
x = np.array([[1, 0, 0], [1, 0, 0], [2, 0, 0]])
with self.cached_session() as sess:
y0, idx0 = gen_array_ops.unique_v2(x, axis=np.array([0], dtype))
tf_y0, tf_idx0 = self.evaluate([y0, idx0])
y1, idx1 = gen_array_ops.unique_v2(x, axis=np.array([1], dtype))
tf_y1, tf_idx1 = self.evaluate([y1, idx1])
self.assertAllEqual(tf_y0, np.array([[1, 0, 0], [2, 0, 0]]))
self.assertAllEqual(tf_idx0, np.array([0, 0, 1]))
self.assertAllEqual(tf_y1, np.array([[1, 0], [1, 0], [2, 0]]))
self.assertAllEqual(tf_idx1, np.array([0, 1, 1]))
def testInt32Axis(self):
for dtype in [np.int32, np.int64]:
x = np.array([[1, 0, 0], [1, 0, 0], [2, 0, 0]])
y0, idx0 = gen_array_ops.unique_v2(x, axis=np.array([0], dtype))
self.assertEqual(y0.shape.rank, 2)
tf_y0, tf_idx0 = self.evaluate([y0, idx0])
y1, idx1 = gen_array_ops.unique_v2(x, axis=np.array([1], dtype))
self.assertEqual(y1.shape.rank, 2)
tf_y1, tf_idx1 = self.evaluate([y1, idx1])
self.assertAllEqual(tf_y0, np.array([[1, 0, 0], [2, 0, 0]]))
self.assertAllEqual(tf_idx0, np.array([0, 0, 1]))
self.assertAllEqual(tf_y1, np.array([[1, 0], [1, 0], [2, 0]]))
self.assertAllEqual(tf_idx1, np.array([0, 1, 1]))
def unique_2d(tensor):
from tensorflow.python.ops import gen_array_ops
ids, idx = gen_array_ops.unique_v2(tensor, axis=[0])
# static shape of unique_v2 ids (0th) and idx(1th) are swapped by mistake.
ids = tf.reshape(ids, [-1, tf.shape(tensor)[1]])
idx = tf.reshape(idx, [-1])
return ids, idx
def testBoolV2(self):
x = np.random.choice([True, False], size=7000)
y, idx = gen_array_ops.unique_v2(x, axis=np.array([], np.int32))
tf_y, tf_idx = self.evaluate([y, idx])
self.assertEqual(len(x), len(tf_idx))
self.assertEqual(len(tf_y), len(np.unique(x)))
for i in range(len(x)):
self.assertEqual(x[i], tf_y[tf_idx[i]])
def testInt32V2(self):
# This test is only temporary, once V2 is used
# by default, the axis will be wrapped to allow `axis=None`.
x = np.random.randint(2, high=10, size=7000)
y, idx = gen_array_ops.unique_v2(x, axis=np.array([], np.int32))
tf_y, tf_idx = self.evaluate([y, idx])
self.assertEqual(len(x), len(tf_idx))
self.assertEqual(len(tf_y), len(np.unique(x)))
for i in range(len(x)):
self.assertEqual(x[i], tf_y[tf_idx[i]])
def testInt32V2(self):
# This test is only temporary, once V2 is used
# by default, the axis will be wrapped to allow `axis=None`.
x = np.random.randint(2, high=10, size=7000)
with self.test_session() as sess:
y, idx = gen_array_ops.unique_v2(x, axis=[])
tf_y, tf_idx = sess.run([y, idx])
self.assertEqual(len(x), len(tf_idx))
self.assertEqual(len(tf_y), len(np.unique(x)))
for i in range(len(x)):
self.assertEqual(x[i], tf_y[tf_idx[i]])
def test_tf_img_to_minmax(self):
# Expected Result
threshold = 127
min_maxed_unique_values = (255, 0)
number_of_black_color = 31760
# 1) Get sample image as tf
test_file_name: str = "sample.png"
test_image_fullpath: str = os.path.join(
self.test_path, self.test_resource_folder_name, test_file_name
)
grayscale_tf_image = tf_images.decode_png(test_image_fullpath, 1)
# 2) Calculate min max
min_maxed_grayscale_tf_image = tf_images.tf_img_to_minmax(
grayscale_tf_image, threshold, (0, 255)
)
# check unique
reshaped_min_maxed_grayscale_tf_image = tf.reshape(
min_maxed_grayscale_tf_image, (-1, 1)
)
unique_min_max_values = gen_array_ops.unique_v2(
reshaped_min_maxed_grayscale_tf_image, axis=[-2]
)[0]
self.assertTrue(
tf.math.reduce_all(
tf.math.equal(
unique_min_max_values,
tf.cast(
tf.expand_dims(tf.constant(min_maxed_unique_values), axis=-1),
tf.float32,
),
)
)
)
self.assertTrue(
tf.math.equal(
tf.math.count_nonzero(min_maxed_grayscale_tf_image),
number_of_black_color,
)
)
def testShapeInferenceV2(self):
"""Test shape inference."""
x = np.arange(6).reshape(3, 2, 1)
_, idx = gen_array_ops.unique_v2(x, axis=[0])
self.assertEqual(idx.shape.as_list(), [3])
_, idx = gen_array_ops.unique_v2(x, axis=[1])
self.assertEqual(idx.shape.as_list(), [2])
_, idx = gen_array_ops.unique_v2(x, axis=[2])
self.assertEqual(idx.shape.as_list(), [1])
_, idx = gen_array_ops.unique_v2(x, axis=[-1])
self.assertEqual(idx.shape.as_list(), [1])
_, idx = gen_array_ops.unique_v2(x, axis=[-2])
self.assertEqual(idx.shape.as_list(), [2])
_, idx = gen_array_ops.unique_v2(x, axis=[-3])
self.assertEqual(idx.shape.as_list(), [3])
_, idx = gen_array_ops.unique_v2([0, 1, 2], axis=[])
self.assertEqual(idx.shape.as_list(), [3])
with self.assertRaisesRegexp(ValueError, "axis expects a 1D vector"):
gen_array_ops.unique_v2(x, axis=[[0]])
with self.assertRaisesRegexp(ValueError, "x expects a 1D vector"):
gen_array_ops.unique_v2(x, axis=[])
with self.assertRaisesRegexp(
ValueError, "axis does not support input tensors larger than"):
gen_array_ops.unique_v2(x, axis=[1, 2])
with self.assertRaisesRegexp(
ValueError, r"axis expects to be in the range \[-3, 3\)"):
gen_array_ops.unique_v2(x, axis=[3])
with self.assertRaisesRegexp(
ValueError, r"axis expects to be in the range \[-3, 3\)"):
gen_array_ops.unique_v2(x, axis=[-4])
x_t = array_ops.placeholder(dtypes.int32, shape=None)
_, idx = gen_array_ops.unique_v2(x_t, axis=[0])
self.assertEqual(idx.shape.as_list(), [None])
axis_t = array_ops.placeholder(dtypes.int32, shape=None)
_, idx = gen_array_ops.unique_v2(x, axis=axis_t)
self.assertEqual(idx.shape.as_list(), [None])
