shape checking in python code (#257)

Author: WindQAQ

tensorflow_addons/image/distort_image_ops.py

@@ -138,6 +138,10 @@ def adjust_hsv_in_yiq(image,
     """
     with tf.name_scope(name or "adjust_hsv_in_yiq"):
         image = tf.convert_to_tensor(image, name="image")
+
+        if image.shape.rank < 3:
+            raise ValueError("input must be at least rank 3.")
+
         # Remember original dtype to so we can convert back if needed
         orig_dtype = image.dtype
         flt_image = tf.image.convert_image_dtype(image, tf.dtypes.float32)

tensorflow_addons/image/distort_image_ops_test.py

@@ -107,7 +107,7 @@ def test_invalid_shapes(self):
         x_np = np.random.rand(2, 3) * 255.
         delta_h = np.random.rand() * 2.0 - 1.0
         with self.assertRaisesRegexp(ValueError,
-                                     "Shape must be at least rank 3"):
+                                     "input must be at least rank 3."):
             self._adjust_hue_in_yiq_tf(x_np, delta_h)
         x_np = np.random.rand(4, 2, 4) * 255.
         delta_h = np.random.rand() * 2.0 - 1.0
@@ -168,7 +168,7 @@ def test_invalid_shapes(self):
         x_np = np.random.rand(2, 3) * 255.
         scale = np.random.rand() * 2.0 - 1.0
         with self.assertRaisesRegexp(ValueError,
-                                     "Shape must be at least rank 3"):
+                                     "input must be at least rank 3."):
             self._adjust_value_in_yiq_tf(x_np, scale)
         x_np = np.random.rand(4, 2, 4) * 255.
         scale = np.random.rand() * 2.0 - 1.0
@@ -233,7 +233,7 @@ def test_invalid_shapes(self):
         x_np = np.random.rand(2, 3) * 255.
         scale = np.random.rand() * 2.0 - 1.0
         with self.assertRaisesRegexp(ValueError,
-                                     "Shape must be at least rank 3"):
+                                     "input must be at least rank 3."):
             self._adjust_saturation_in_yiq_tf(x_np, scale)
         x_np = np.random.rand(4, 2, 4) * 255.
         scale = np.random.rand() * 2.0 - 1.0