Fix bug in "xy" mode for bilinear interpolation (#845)

* Fix bug in "xy" mode for bilinear interpolation
* fixed error in interp implementation, and added test for non-square images.
* removed erroneous non-square test, and modified small_grid tests to be non-square.

Author: djl11

tensorflow_addons/image/dense_image_warp.py

@@ -116,11 +116,11 @@ def interpolate_bilinear(grid, query_points, indexing="ij", name=None):
         index_order = [0, 1] if indexing == "ij" else [1, 0]
         unstacked_query_points = tf.unstack(query_points, axis=2, num=2)
 
-        for dim in index_order:
+        for i, dim in enumerate(index_order):
             with tf.name_scope("dim-" + str(dim)):
                 queries = unstacked_query_points[dim]
 
-                size_in_indexing_dimension = grid_shape[dim + 1]
+                size_in_indexing_dimension = grid_shape[i + 1]
 
                 # max_floor is size_in_indexing_dimension - 2 so that max_floor + 1
                 # is still a valid index into the grid.

tensorflow_addons/image/dense_image_warp_test.py

@@ -30,22 +30,28 @@
 @test_utils.run_all_in_graph_and_eager_modes
 class InterpolateBilinearTest(tf.test.TestCase):
     def test_interpolate_small_grid_ij(self):
-        grid = tf.constant([[0., 1., 2.], [3., 4., 5.], [6., 7., 8.]],
-                           shape=[1, 3, 3, 1])
-        query_points = tf.constant([[0., 0.], [1., 0.], [2., 0.5], [1.5, 1.5]],
-                                   shape=[1, 4, 2])
-        expected_results = np.reshape(np.array([0., 3., 6.5, 6.]), [1, 4, 1])
+        grid = tf.constant(
+            [[0., 1., 2.], [3., 4., 5.], [6., 7., 8.], [9., 10., 11.]],
+            shape=[1, 4, 3, 1])
+        query_points = tf.constant(
+            [[0., 0.], [1., 0.], [2., 0.5], [1.5, 1.5], [3., 2.]],
+            shape=[1, 5, 2])
+        expected_results = np.reshape(
+            np.array([0., 3., 6.5, 6., 11.]), [1, 5, 1])
 
         interp = interpolate_bilinear(grid, query_points)
 
         self.assertAllClose(expected_results, interp)
 
     def test_interpolate_small_grid_xy(self):
-        grid = tf.constant([[0., 1., 2.], [3., 4., 5.], [6., 7., 8.]],
-                           shape=[1, 3, 3, 1])
+        grid = tf.constant(
+            [[0., 1., 2.], [3., 4., 5.], [6., 7., 8.], [9., 10., 11.]],
+            shape=[1, 4, 3, 1])
         query_points = tf.constant(
-            [[0., 0.], [0., 1.], [0.5, 2.0], [1.5, 1.5]], shape=[1, 4, 2])
-        expected_results = np.reshape(np.array([0., 3., 6.5, 6.]), [1, 4, 1])
+            [[0., 0.], [0., 1.], [0.5, 2.0], [1.5, 1.5], [2., 3.]],
+            shape=[1, 5, 2])
+        expected_results = np.reshape(
+            np.array([0., 3., 6.5, 6., 11.]), [1, 5, 1])
 
         interp = interpolate_bilinear(grid, query_points, indexing="xy")