No public description

PiperOrigin-RevId: 689038687

official/vision/ops/preprocess_ops.py

@@ -182,7 +182,7 @@ def resize_and_crop_image(
   2. Pad the rescaled image to the padded_size.
 
   Args:
-    image: a `Tensor` of shape [height, width, 3] representing an image.
+    image: a `Tensor` of shape [height, width, c] representing an image.
     desired_size: a `Tensor` or `int` list/tuple of two elements representing
       [height, width] of the desired actual output image size.
     padded_size: a `Tensor` or `int` list/tuple of two elements representing
@@ -201,7 +201,7 @@ def resize_and_crop_image(
       behaviour is to place it at left top corner.
 
   Returns:
-    output_image: `Tensor` of shape [height, width, 3] where [height, width]
+    output_image: `Tensor` of shape [height, width, c] where [height, width]
       equals to `output_size`.
     image_info: a 2D `Tensor` that encodes the information of the image and the
       applied preprocessing. It is in the format of
@@ -786,7 +786,7 @@ def random_horizontal_flip(
   """Randomly flips input image and bounding boxes and/or masks horizontally.
 
   Expects input tensors without the batch dimension; i.e. for RGB image assume
-  rank-3 input like [h, w, 3], for masks assume either [h, w, 1] or [1, h, w].
+  rank-3 input like [h, w, c], for masks assume either [h, w, 1] or [1, h, w].
 
   Args:
     image: `tf.Tensor`, the image to apply the random flip, [h, w, channels].

official/vision/ops/preprocess_ops_test.py

@@ -80,11 +80,19 @@ def test_pad_to_fixed_size(
           aug_scale_max=2.0,
           output_scales=(20 / 100, 10 / 200),
       ),
+      dict(
+          testcase_name='no_jittering_with_4_channels',
+          input_size=(100, 200),
+          desired_size=(20, 10),
+          aug_scale_max=1.0,
+          output_scales=(20 / 100, 10 / 200),
+          channels=4,
+      ),
   )
   def test_resize_and_crop_image_not_keep_aspect_ratio(
-      self, input_size, desired_size, aug_scale_max, output_scales
+      self, input_size, desired_size, aug_scale_max, output_scales, channels=3
   ):
-    image = tf.convert_to_tensor(np.random.rand(*input_size, 3))
+    image = tf.convert_to_tensor(np.random.rand(*input_size, channels))
 
     resized_image, image_info = preprocess_ops.resize_and_crop_image(
         image,
@@ -95,7 +103,7 @@ def test_resize_and_crop_image_not_keep_aspect_ratio(
     )
     resized_image_shape = tf.shape(resized_image)
 
-    self.assertAllEqual([*desired_size, 3], resized_image_shape.numpy())
+    self.assertAllEqual([*desired_size, channels], resized_image_shape.numpy())
     if aug_scale_max == 1:
       self.assertNDArrayNear(
           [input_size, desired_size, output_scales, [0.0, 0.0]],
@@ -108,7 +116,7 @@ def test_resize_and_crop_image_not_keep_aspect_ratio(
       (100, 256, 128, 256, 32, 1.0, 1.0, 128, 256),
       (200, 512, 200, 128, 32, 0.25, 0.25, 224, 128),
   )
-  def test_resize_and_crop_image_rectangluar_case(
+  def test_resize_and_crop_image_rectangular_case(
       self,
       input_height,
       input_width,