Merge pull request #3293 from pkulzc/master

Internal changes of object_detection 

research/object_detection/core/matcher_test.py

@@ -20,7 +20,7 @@ import tensorflow as tf
 from object_detection.core import matcher
 
 
-class AnchorMatcherTest(tf.test.TestCase):
+class MatchTest(tf.test.TestCase):
 
   def test_get_correct_matched_columnIndices(self):
     match_results = tf.constant([3, 1, -1, 0, -1, 5, -2])
@@ -145,6 +145,48 @@ class AnchorMatcherTest(tf.test.TestCase):
       self.assertAllEqual(all_indices_sorted,
                           np.arange(num_matches, dtype=np.int32))
 
+  def test_scalar_gather_based_on_match(self):
+    match_results = tf.constant([3, 1, -1, 0, -1, 5, -2])
+    input_tensor = tf.constant([0, 1, 2, 3, 4, 5, 6, 7], dtype=tf.float32)
+    expected_gathered_tensor = [3, 1, 100, 0, 100, 5, 200]
+    match = matcher.Match(match_results)
+    gathered_tensor = match.gather_based_on_match(input_tensor,
+                                                  unmatched_value=100.,
+                                                  ignored_value=200.)
+    self.assertEquals(gathered_tensor.dtype, tf.float32)
+    with self.test_session():
+      gathered_tensor_out = gathered_tensor.eval()
+    self.assertAllEqual(expected_gathered_tensor, gathered_tensor_out)
+
+  def test_multidimensional_gather_based_on_match(self):
+    match_results = tf.constant([1, -1, -2])
+    input_tensor = tf.constant([[0, 0.5, 0, 0.5], [0, 0, 0.5, 0.5]],
+                               dtype=tf.float32)
+    expected_gathered_tensor = [[0, 0, 0.5, 0.5], [0, 0, 0, 0], [0, 0, 0, 0]]
+    match = matcher.Match(match_results)
+    gathered_tensor = match.gather_based_on_match(input_tensor,
+                                                  unmatched_value=tf.zeros(4),
+                                                  ignored_value=tf.zeros(4))
+    self.assertEquals(gathered_tensor.dtype, tf.float32)
+    with self.test_session():
+      gathered_tensor_out = gathered_tensor.eval()
+    self.assertAllEqual(expected_gathered_tensor, gathered_tensor_out)
+
+  def test_multidimensional_gather_based_on_match_with_matmul_gather_op(self):
+    match_results = tf.constant([1, -1, -2])
+    input_tensor = tf.constant([[0, 0.5, 0, 0.5], [0, 0, 0.5, 0.5]],
+                               dtype=tf.float32)
+    expected_gathered_tensor = [[0, 0, 0.5, 0.5], [0, 0, 0, 0], [0, 0, 0, 0]]
+    match = matcher.Match(match_results, use_matmul_gather=True)
+    gathered_tensor = match.gather_based_on_match(input_tensor,
+                                                  unmatched_value=tf.zeros(4),
+                                                  ignored_value=tf.zeros(4))
+    self.assertEquals(gathered_tensor.dtype, tf.float32)
+    with self.test_session() as sess:
+      self.assertTrue(
+          all([op.name is not 'Gather' for op in sess.graph.get_operations()]))
+      gathered_tensor_out = gathered_tensor.eval()
+    self.assertAllEqual(expected_gathered_tensor, gathered_tensor_out)
 
 if __name__ == '__main__':
   tf.test.main()

research/object_detection/core/standard_fields.py

@@ -57,6 +57,10 @@ class InputDataFields(object):
     groundtruth_keypoints: ground truth keypoints.
     groundtruth_keypoint_visibilities: ground truth keypoint visibilities.
     groundtruth_label_scores: groundtruth label scores.
+    groundtruth_weights: groundtruth weight factor for bounding boxes.
+    num_groundtruth_boxes: number of groundtruth boxes.
+    true_image_shapes: true shapes of images in the resized images, as resized
+      images can be padded with zeros.
   """
   image = 'image'
   original_image = 'original_image'
@@ -79,10 +83,13 @@ class InputDataFields(object):
   groundtruth_keypoints = 'groundtruth_keypoints'
   groundtruth_keypoint_visibilities = 'groundtruth_keypoint_visibilities'
   groundtruth_label_scores = 'groundtruth_label_scores'
+  groundtruth_weights = 'groundtruth_weights'
+  num_groundtruth_boxes = 'num_groundtruth_boxes'
+  true_image_shape = 'true_image_shape'
 
 
 class DetectionResultFields(object):
-  """Naming converntions for storing the output of the detector.
+  """Naming conventions for storing the output of the detector.
 
   Attributes:
     source_id: source of the original image.
@@ -162,6 +169,7 @@ class TfExampleFields(object):
     object_is_crowd: [DEPRECATED, use object_group_of instead]
       is the object a single object or a crowd
     object_segment_area: the area of the segment.
+    object_weight: a weight factor for the object's bounding box.
     instance_masks: instance segmentation masks.
     instance_boundaries: instance boundaries.
     instance_classes: Classes for each instance segmentation mask.
@@ -194,6 +202,7 @@ class TfExampleFields(object):
   object_depiction = 'image/object/depiction'
   object_is_crowd = 'image/object/is_crowd'
   object_segment_area = 'image/object/segment/area'
+  object_weight = 'image/object/weight'
   instance_masks = 'image/segmentation/object'
   instance_boundaries = 'image/boundaries/object'
   instance_classes = 'image/segmentation/object/class'

research/object_detection/dataset_tools/__init__.py

-