Small polishing changes to DELF code.

PiperOrigin-RevId: 185469605

research/delf/delf/python/datum_io.py

@@ -24,10 +24,11 @@
 from __future__ import division
 from __future__ import print_function
 
-from delf import datum_pb2
 import numpy as np
 import tensorflow as tf
 
+from delf import datum_pb2
+
 
 def ArrayToDatum(arr):
   """Converts numpy array to DatumProto.

research/delf/delf/python/datum_io_test.py

@@ -12,19 +12,19 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 # ==============================================================================
-
 """Tests for datum_io, the python interface of DatumProto."""
 
 from __future__ import absolute_import
 from __future__ import division
 from __future__ import print_function
 
-from delf import datum_io
-from delf import datum_pb2
-import numpy as np
 import os
+
+import numpy as np
 import tensorflow as tf
 
+from delf import datum_io
+
 
 class DatumIoTest(tf.test.TestCase):
 

research/delf/delf/python/delf_v1.py

@@ -12,8 +12,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 # ==============================================================================
-
-"""DELF model implementation based on the following paper:
+"""DELF model implementation based on the following paper.
 
   Large-Scale Image Retrieval with Attentive Deep Local Features
   https://arxiv.org/abs/1612.06321
@@ -26,10 +25,10 @@
 from __future__ import division
 from __future__ import print_function
 
-from nets import resnet_v1
-
 import tensorflow as tf
 
+from nets import resnet_v1
+
 slim = tf.contrib.slim
 
 _SUPPORTED_TARGET_LAYER = ['resnet_v1_50/block3', 'resnet_v1_50/block4']

research/delf/delf/python/examples/extract_features.py

@@ -12,7 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 # ==============================================================================
-
 """Extracts DELF features from a list of images, saving them to file.
 
 The images must be in JPG format. The program checks if descriptors already
@@ -24,18 +23,17 @@
 from __future__ import print_function
 
 import argparse
-from google.protobuf import text_format
-import numpy as np
 import os
 import sys
-import tensorflow as tf
-from tensorflow.python.platform import app
 import time
 
+import tensorflow as tf
+
+from google.protobuf import text_format
+from tensorflow.python.platform import app
 from delf import delf_config_pb2
 from delf import feature_extractor
 from delf import feature_io
-from delf import feature_pb2
 
 cmd_args = None
 
@@ -152,9 +150,9 @@ def main(unused_argv):
                      config.delf_local_config.max_feature_num
              })
 
-        serialized_desc = feature_io.WriteToFile(
-            out_desc_fullpath, locations_out, feature_scales_out,
-            descriptors_out, attention_out)
+        feature_io.WriteToFile(out_desc_fullpath, locations_out,
+                               feature_scales_out, descriptors_out,
+                               attention_out)
 
       # Finalize enqueue threads.
       coord.request_stop()

research/delf/delf/python/examples/match_images.py

@@ -12,7 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 # ==============================================================================
-
 """Matches two images using their DELF features.
 
 The matching is done using feature-based nearest-neighbor search, followed by
@@ -26,17 +25,19 @@
 from __future__ import print_function
 
 import argparse
-from delf import feature_io
+import sys
+
 import matplotlib.image as mpimg
 import matplotlib.pyplot as plt
 import numpy as np
 from scipy.spatial import cKDTree
 from skimage.feature import plot_matches
 from skimage.measure import ransac
 from skimage.transform import AffineTransform
-import sys
 import tensorflow as tf
+
 from tensorflow.python.platform import app
+from delf import feature_io
 
 cmd_args = None
 
@@ -58,21 +59,23 @@ def main(unused_argv):
 
   # Find nearest-neighbor matches using a KD tree.
   d1_tree = cKDTree(descriptors_1)
-  distances, indices = d1_tree.query(
+  _, indices = d1_tree.query(
       descriptors_2, distance_upper_bound=_DISTANCE_THRESHOLD)
 
   # Select feature locations for putative matches.
   locations_2_to_use = np.array([
-      locations_2[i,] for i in range(num_features_2)
+      locations_2[i,]
+      for i in range(num_features_2)
       if indices[i] != num_features_1
   ])
   locations_1_to_use = np.array([
-      locations_1[indices[i],] for i in range(num_features_2)
+      locations_1[indices[i],]
+      for i in range(num_features_2)
       if indices[i] != num_features_1
   ])
 
   # Perform geometric verification using RANSAC.
-  model_robust, inliers = ransac(
+  _, inliers = ransac(
       (locations_1_to_use, locations_2_to_use),
       AffineTransform,
       min_samples=3,
@@ -82,7 +85,7 @@ def main(unused_argv):
   tf.logging.info('Found %d inliers' % sum(inliers))
 
   # Visualize correspondences, and save to file.
-  fig, ax = plt.subplots()
+  _, ax = plt.subplots()
   img_1 = mpimg.imread(cmd_args.image_1_path)
   img_2 = mpimg.imread(cmd_args.image_2_path)
   inlier_idxs = np.nonzero(inliers)[0]

research/delf/delf/python/feature_extractor.py

@@ -18,12 +18,12 @@
 from __future__ import division
 from __future__ import print_function
 
+import tensorflow as tf
+
 from delf import datum_io
 from delf import delf_v1
-from delf import delf_config_pb2
 from object_detection.core import box_list
 from object_detection.core import box_list_ops
-import tensorflow as tf
 
 
 def NormalizePixelValues(image,
@@ -170,7 +170,8 @@ def _ProcessSingleScale(scale_index,
         resized_image, normalized_image=True, reuse=reuse)
 
     rf_boxes = CalculateReceptiveBoxes(
-        tf.shape(feature_map)[1], tf.shape(feature_map)[2], rf, stride, padding)
+        tf.shape(feature_map)[1],
+        tf.shape(feature_map)[2], rf, stride, padding)
     # Re-project back to the original image space.
     rf_boxes = tf.divide(rf_boxes, scale)
     attention = tf.reshape(attention, [-1])
@@ -236,8 +237,8 @@ def _ProcessSingleScale(scale_index,
                                                  nms_max_boxes)
 
   return (final_boxes.get(), final_boxes.get_field('scales'),
-          final_boxes.get_field('features'), tf.expand_dims(
-              final_boxes.get_field('scores'), 1))
+          final_boxes.get_field('features'),
+          tf.expand_dims(final_boxes.get_field('scores'), 1))
 
 
 def BuildModel(layer_name, attention_nonlinear, attention_type,

research/delf/delf/python/feature_extractor_test.py

@@ -12,17 +12,17 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 # ==============================================================================
-
 """Tests for DELF feature extractor."""
 
 from __future__ import absolute_import
 from __future__ import division
 from __future__ import print_function
 
-from delf import feature_extractor
 import numpy as np
 import tensorflow as tf
 
+from delf import feature_extractor
+
 
 class FeatureExtractorTest(tf.test.TestCase):
 
@@ -80,14 +80,15 @@ def _test_model_fn(image, normalized_image, reuse):
           axis=3)
       return attention, feature_map
 
-    boxes, feature_scales, features, scores = feature_extractor.ExtractKeypointDescriptor(
-        image,
-        layer_name='resnet_v1_50/block3',
-        image_scales=tf.constant([1.0]),
-        iou=1.0,
-        max_feature_num=10,
-        abs_thres=1.5,
-        model_fn=_test_model_fn)
+    boxes, feature_scales, features, scores = (
+        feature_extractor.ExtractKeypointDescriptor(
+            image,
+            layer_name='resnet_v1_50/block3',
+            image_scales=tf.constant([1.0]),
+            iou=1.0,
+            max_feature_num=10,
+            abs_thres=1.5,
+            model_fn=_test_model_fn))
 
     exp_boxes = [[-145.0, -145.0, 145.0, 145.0], [-113.0, -145.0, 177.0, 145.0]]
     exp_feature_scales = [1.0, 1.0]

research/delf/delf/python/feature_io.py

@@ -21,11 +21,12 @@
 from __future__ import division
 from __future__ import print_function
 
-from delf import feature_pb2
-from delf import datum_io
 import numpy as np
 import tensorflow as tf
 
+from delf import feature_pb2
+from delf import datum_io
+
 
 def ArraysToDelfFeatures(locations,
                          scales,
@@ -57,7 +58,7 @@ def ArraysToDelfFeatures(locations,
     assert num_features == len(orientations)
 
   delf_features = feature_pb2.DelfFeatures()
-  for i in xrange(num_features):
+  for i in range(num_features):
     delf_feature = delf_features.feature.add()
     delf_feature.y = locations[i, 0]
     delf_feature.x = locations[i, 1]
@@ -98,7 +99,7 @@ def DelfFeaturesToArrays(delf_features):
   attention = np.zeros([num_features])
   orientations = np.zeros([num_features])
 
-  for i in xrange(num_features):
+  for i in range(num_features):
     delf_feature = delf_features.feature[i]
     locations[i, 0] = delf_feature.y
     locations[i, 1] = delf_feature.x

research/delf/delf/python/feature_io_test.py

@@ -12,30 +12,31 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 # ==============================================================================
-
 """Tests for feature_io, the python interface of DelfFeatures."""
 
 from __future__ import absolute_import
 from __future__ import division
 from __future__ import print_function
 
-from delf import feature_io
-from delf import datum_pb2
-import numpy as np
 import os
+
+import numpy as np
 import tensorflow as tf
 
+from delf import feature_io
+
 
 def create_data():
   """Creates data to be used in tests.
-    Returns:
-      locations: [N, 2] float array which denotes the selected keypoint
-        locations. N is the number of features.
-      scales: [N] float array with feature scales.
-      descriptors: [N, depth] float array with DELF descriptors.
-      attention: [N] float array with attention scores.
-      orientations: [N] float array with orientations.
-    """
+
+  Returns:
+    locations: [N, 2] float array which denotes the selected keypoint
+      locations. N is the number of features.
+    scales: [N] float array with feature scales.
+    descriptors: [N, depth] float array with DELF descriptors.
+    attention: [N] float array with attention scores.
+    orientations: [N] float array with orientations.
+  """
   locations = np.arange(8, dtype=np.float32).reshape(4, 2)
   scales = np.arange(4, dtype=np.float32)
   attention = np.arange(4, dtype=np.float32)