DELF codebase general cleanup (#9930)

* Merged commit includes the following changes:
253126424  by Andre Araujo:

    Scripts to compute metrics for Google Landmarks dataset.

    Also, a small fix to metric in retrieval case: avoids duplicate predicted images.

--
253118971  by Andre Araujo:

    Metrics for Google Landmarks dataset.

--
253106953  by Andre Araujo:

    Library to read files from Google Landmarks challenges.

--
250700636  by Andre Araujo:

    Handle case of aggregation extraction with empty set of input features.

--
250516819  by Andre Araujo:

    Add minimum size for DELF extractor.

--
250435822  by Andre Araujo:

    Add max_image_size/min_image_size for open-source DELF proto / module.

--
250414606  by Andre Araujo:

    Refactor extract_aggregation to allow reuse with different datasets.

--
250356863  by Andre Araujo:

    Remove unnecessary cmd_args variable from boxes_and_features_extraction.

--
249783379  by Andre Araujo:

    Create directory for writing mapping file if it does not exist.

--
249581591  by Andre Araujo:

    Refactor scripts to extract boxes and features from images in Revisited datasets.
    Also, change tf.logging.info --> print for easier logging in open source code.

--
249511821  by Andre Araujo:

    Small change to function for file/directory handling.

--
249289499  by Andre Araujo:

    Internal change.

--

PiperOrigin-RevId: 253126424

* Updating DELF init to adjust to latest changes

* Editing init files for python packages

* Edit D2R dataset reader to work with py3.

PiperOrigin-RevId: 253135576

* DELF package: fix import ordering

* Adding new requirements to setup.py

* Adding init file for training dir

* Merged commit includes the following changes:

FolderOrigin-RevId: /google/src/cloud/andrearaujo/delf_oss/google3/..

* Adding init file for training subdirs

* Working version of DELF training

* Internal change.

PiperOrigin-RevId: 253248648

* Fix variance loading in open-source code.

PiperOrigin-RevId: 260619120

* Separate image re-ranking as a standalone library, and add metric writing to dataset library.

PiperOrigin-RevId: 260998608

* Tool to read written D2R Revisited datasets metrics file. Test is added.

Also adds a unit test for previously-existing SaveMetricsFile function.

PiperOrigin-RevId: 263361410

* Add optional resize factor for feature extraction.

PiperOrigin-RevId: 264437080

* Fix NumPy's new version spacing changes.

PiperOrigin-RevId: 265127245

* Maker image matching function visible, and add support for RANSAC seed.

PiperOrigin-RevId: 277177468

* Avoid matplotlib failure due to missing display backend.

PiperOrigin-RevId: 287316435

* Removes tf.contrib dependency.

PiperOrigin-RevId: 288842237

* Fix tf contrib removal for feature_aggregation_extractor.

PiperOrigin-RevId: 289487669

* Merged commit includes the following changes:
309118395  by Andre Araujo:

    Make DELF open-source code compatible with TF2.

--
309067582  by Andre Araujo:

    Handle image resizing rounding properly for python extraction.

    New behavior is tested with unit tests.

--
308690144  by Andre Araujo:

    Several changes to improve DELF model/training code and make it work in TF 2.1.0:
    - Rename some files for better clarity
    - Using compat.v1 versions of functions
    - Formatting changes
    - Using more appropriate TF function names

--
308689397  by Andre Araujo:

    Internal change.

--
308341315  by Andre Araujo:

    Remove old slim dependency in DELF open-source model.

    This avoids issues with requiring old TF-v1, making it compatible with latest TF.

--
306777559  by Andre Araujo:

    Internal change

--
304505811  by Andre Araujo:

    Raise error during geometric verification if local features have different dimensionalities.

--
301739992  by Andre Araujo:

    Transform some geometric verification constants into arguments, to allow custom matching.

--
301300324  by Andre Araujo:

    Apply name change(experimental_run_v2 -> run) for all callers in Tensorflow.

--
299919057  by Andre Araujo:

    Automated refactoring to make code Python 3 compatible.

--
297953698  by Andre Araujo:

    Explicitly replace "import tensorflow" with "tensorflow.compat.v1" for TF2.x migration

--
297521242  by Andre Araujo:

    Explicitly replace "import tensorflow" with "tensorflow.compat.v1" for TF2.x migration

--
297278247  by Andre Araujo:

    Explicitly replace "import tensorflow" with "tensorflow.compat.v1" for TF2.x migration

--
297270405  by Andre Araujo:

    Explicitly replace "import tensorflow" with "tensorflow.compat.v1" for TF2.x migration

--
297238741  by Andre Araujo:

    Explicitly replace "import tensorflow" with "tensorflow.compat.v1" for TF2.x migration

--
297108605  by Andre Araujo:

    Explicitly replace "import tensorflow" with "tensorflow.compat.v1" for TF2.x migration

--
294676131  by Andre Araujo:

    Add option to resize images to square resolutions without aspect ratio preservation.

--
293849641  by Andre Araujo:

    Internal change.

--
293840896  by Andre Araujo:

    Changing Slim import to tf_slim codebase.

--
293661660  by Andre Araujo:

    Allow the delf training script to read from TFRecords dataset.

--
291755295  by Andre Araujo:

    Internal change.

--
291448508  by Andre Araujo:

    Internal change.

--
291414459  by Andre Araujo:

    Adding train script.

--
291384336  by Andre Araujo:

    Adding model export script and test.

--
291260565  by Andre Araujo:

    Adding placeholder for Google Landmarks dataset.

--
291205548  by Andre Araujo:

    Definition of DELF model using Keras ResNet50 as backbone.

--
289500793  by Andre Araujo:

    Add TFRecord building script for delf.

--

PiperOrigin-RevId: 309118395

* Updating README, dependency versions

* Updating training README

* Fixing init import of export_model

* Fixing init import of export_model_utils

* tkinter in INSTALL_INSTRUCTIONS

* Merged commit includes the following changes:

FolderOrigin-RevId: /google/src/cloud/andrearaujo/delf_oss/google3/..

* INSTALL_INSTRUCTIONS mentioning different cloning options

* Updating required TF version, since 2.1 is not available in pip

* Internal change.

PiperOrigin-RevId: 309136003

* Fix missing string_input_producer and start_queue_runners in TF2.

PiperOrigin-RevId: 309437512

* Handle RANSAC from skimage's latest versions.

PiperOrigin-RevId: 310170897

* DELF 2.1 version: badge and setup.py updated

* Add TF version badge in INSTALL_INSTRUCTIONS and paper badges in README

* Add paper badges in paper instructions

* Add paper badge to landmark detection instructions

* Small update to DELF training README

* Merged commit includes the following changes:
312614961  by Andre Araujo:

    Instructions/code to reproduce DELG paper results.

--
312523414  by Andre Araujo:

    Fix a minor bug when post-process extracted features, format config.delf_global_config.image_scales_ind to a list.

--
312340276  by Andre Araujo:

    Add support for global feature extraction in DELF open-source codebase.

--
311031367  by Andre Araujo:

    Add use_square_images as an option in DELF config. The default value is false. if it is set, then images are resized to square resolution before feature extraction (e.g. Starburst use case. ) Thought for a while, whether to have two constructor of DescriptorToImageTemplate, but in the end, decide to only keep one, may be less confusing.

--
310658638  by Andre Araujo:

    Option for producing local feature-based image match visualization.

--

PiperOrigin-RevId: 312614961

* DELF README update / DELG instructions

* DELF README update

* DELG instructions update

* Merged commit includes the following changes:

PiperOrigin-RevId: 312695597

* Merged commit includes the following changes:
312754894  by Andre Araujo:

    Code edits / instructions to reproduce GLDv2 results.

--

PiperOrigin-RevId: 312754894

* Markdown updates after adding GLDv2 stuff

* Small updates to DELF README

* Clarify that library must be installed before reproducing results

* Merged commit includes the following changes:
319114828  by Andre Araujo:

    Upgrade global feature model exporting to TF2.

--

PiperOrigin-RevId: 319114828

* Properly merging README

* small edits to README

* small edits to README

* small edits to README

* global feature exporting in training README

* Update to DELF README, install instructions

* Centralizing installation instructions

* Small readme update

* Fixing commas

* Mention DELG acceptance into ECCV'20

* Merged commit includes the following changes:
326723075  by Andre Araujo:

    Move image resize utility into utils.py.

--

PiperOrigin-RevId: 326723075

* Adding back matched_images_demo.png

* Merged commit includes the following changes:
327279047  by Andre Araujo:

    Adapt extractor to handle new form of joint local+global extraction.

--
326733524  by Andre Araujo:

    Internal change.

--

PiperOrigin-RevId: 327279047

* Updated DELG instructions after model extraction refactoring

* Updating GLDv2 paper model baseline

* Merged commit includes the following changes:
328982978  by Andre Araujo:

    Updated DELG model training so that the size of the output tensor is unchanged by the GeM pooling layer. Export global model trained with DELG global features.

--
328218938  by Andre Araujo:

    Internal change.

--

PiperOrigin-RevId: 328982978

* Updated training README after recent changes

* Updated training README to fix small typo

* Merged commit includes the following changes:
330022709  by Andre Araujo:

    Export joint local+global TF2 DELG model, and enable such joint extraction.

    Also, rename export_model.py -> export_local_model.py for better clarity.

    To check that the new exporting code is doing the right thing, I compared features extracted from the new exported model against those extracted from models exported with a single modality, using the same checkpoint. They are identical.

    Some other small changes:
    - small automatic reformating
    - small documentation improvements

--

PiperOrigin-RevId: 330022709

* Updated DELG exporting instructions

* Updated DELG exporting instructions: fix small typo

* Adding DELG pre-trained models on GLDv2-clean

* Merged commit includes the following changes:
331625297  by Andre Araujo:

    Internal change.

--
330062115  by Andre Araujo:

    Fix small (non-critical) typo in the DELG extractor.

--

PiperOrigin-RevId: 331625297

* Merged commit includes the following changes:
347479009  by Andre Araujo:

    Fix image size setting for GLD training.

--

PiperOrigin-RevId: 347479009

* Merged commit includes the following changes:

FolderOrigin-RevId: /google/src/cloud/andrearaujo/copybara_25C283E7A3474256A7C206FC5ABF7E8D_0/google3/..

* Merged commit includes the following changes:

FolderOrigin-RevId: /google/src/cloud/andrearaujo/copybara_25C283E7A3474256A7C206FC5ABF7E8D_0/google3/..

* Merged commit includes the following changes:

FolderOrigin-RevId: /google/src/cloud/andrearaujo/copybara_25C283E7A3474256A7C206FC5ABF7E8D_1/google3/..

* Add whiten module import

research/delf/delf/__init__.py

@@ -30,6 +30,7 @@ from delf.python import feature_aggregation_similarity
 from delf.python import feature_extractor
 from delf.python import feature_io
 from delf.python import utils
+from delf.python import whiten
 from delf.python.examples import detector
 from delf.python.examples import extractor
 from delf.python import detect_to_retrieve

research/delf/delf/python/datasets/google_landmarks_dataset/compute_recognition_metrics.py

@@ -24,7 +24,7 @@ from __future__ import print_function
 import argparse
 import sys
 
-from tensorflow.python.platform import app
+from absl import app
 from delf.python.datasets.google_landmarks_dataset import dataset_file_io
 from delf.python.datasets.google_landmarks_dataset import metrics
 

research/delf/delf/python/datasets/google_landmarks_dataset/compute_retrieval_metrics.py

@@ -24,7 +24,7 @@ from __future__ import print_function
 import argparse
 import sys
 
-from tensorflow.python.platform import app
+from absl import app
 from delf.python.datasets.google_landmarks_dataset import dataset_file_io
 from delf.python.datasets.google_landmarks_dataset import metrics
 

research/delf/delf/python/datasets/google_landmarks_dataset/dataset_file_io_test.py

@@ -32,8 +32,7 @@ class DatasetFileIoTest(tf.test.TestCase):
 
   def testReadRecognitionSolutionWorks(self):
     # Define inputs.
-    file_path = os.path.join(FLAGS.test_tmpdir,
-                             'recognition_solution.csv')
+    file_path = os.path.join(FLAGS.test_tmpdir, 'recognition_solution.csv')
     with tf.io.gfile.GFile(file_path, 'w') as f:
       f.write('id,landmarks,Usage\n')
       f.write('0123456789abcdef,0 12,Public\n')
@@ -64,8 +63,7 @@ class DatasetFileIoTest(tf.test.TestCase):
 
   def testReadRetrievalSolutionWorks(self):
     # Define inputs.
-    file_path = os.path.join(FLAGS.test_tmpdir,
-                             'retrieval_solution.csv')
+    file_path = os.path.join(FLAGS.test_tmpdir, 'retrieval_solution.csv')
     with tf.io.gfile.GFile(file_path, 'w') as f:
       f.write('id,images,Usage\n')
       f.write('0123456789abcdef,None,Ignored\n')
@@ -96,8 +94,7 @@ class DatasetFileIoTest(tf.test.TestCase):
 
   def testReadRecognitionPredictionsWorks(self):
     # Define inputs.
-    file_path = os.path.join(FLAGS.test_tmpdir,
-                             'recognition_predictions.csv')
+    file_path = os.path.join(FLAGS.test_tmpdir, 'recognition_predictions.csv')
     with tf.io.gfile.GFile(file_path, 'w') as f:
       f.write('id,landmarks\n')
       f.write('0123456789abcdef,12 0.1 \n')
@@ -134,8 +131,7 @@ class DatasetFileIoTest(tf.test.TestCase):
 
   def testReadRetrievalPredictionsWorks(self):
     # Define inputs.
-    file_path = os.path.join(FLAGS.test_tmpdir,
-                             'retrieval_predictions.csv')
+    file_path = os.path.join(FLAGS.test_tmpdir, 'retrieval_predictions.csv')
     with tf.io.gfile.GFile(file_path, 'w') as f:
       f.write('id,images\n')
       f.write('0123456789abcdef,fedcba9876543250 \n')

research/delf/delf/python/datasets/revisited_op/dataset.py

@@ -27,6 +27,8 @@ import tensorflow as tf
 
 _GROUND_TRUTH_KEYS = ['easy', 'hard', 'junk']
 
+DATASET_NAMES = ['roxford5k', 'rparis6k']
+
 
 def ReadDatasetFile(dataset_file_path):
   """Reads dataset file in Revisited Oxford/Paris ".mat" format.
@@ -105,14 +107,14 @@ def ParseEasyMediumHardGroundTruth(ground_truth):
   hard_ground_truth = []
   for i in range(num_queries):
     easy_ground_truth.append(
-      _ParseGroundTruth([ground_truth[i]['easy']],
-                        [ground_truth[i]['junk'], ground_truth[i]['hard']]))
+        _ParseGroundTruth([ground_truth[i]['easy']],
+                          [ground_truth[i]['junk'], ground_truth[i]['hard']]))
     medium_ground_truth.append(
-      _ParseGroundTruth([ground_truth[i]['easy'], ground_truth[i]['hard']],
-                        [ground_truth[i]['junk']]))
+        _ParseGroundTruth([ground_truth[i]['easy'], ground_truth[i]['hard']],
+                          [ground_truth[i]['junk']]))
     hard_ground_truth.append(
-      _ParseGroundTruth([ground_truth[i]['hard']],
-                        [ground_truth[i]['junk'], ground_truth[i]['easy']]))
+        _ParseGroundTruth([ground_truth[i]['hard']],
+                          [ground_truth[i]['junk'], ground_truth[i]['easy']]))
 
   return easy_ground_truth, medium_ground_truth, hard_ground_truth
 
@@ -216,13 +218,13 @@ def ComputePRAtRanks(positive_ranks, desired_pr_ranks):
   positive_ranks_one_indexed = positive_ranks + 1
   for i, desired_pr_rank in enumerate(desired_pr_ranks):
     recalls[i] = np.sum(
-      positive_ranks_one_indexed <= desired_pr_rank) / num_expected_positives
+        positive_ranks_one_indexed <= desired_pr_rank) / num_expected_positives
 
     # If `desired_pr_rank` is larger than last positive's rank, only compute
     # precision with respect to last positive's position.
     precision_rank = min(max(positive_ranks_one_indexed), desired_pr_rank)
     precisions[i] = np.sum(
-      positive_ranks_one_indexed <= precision_rank) / precision_rank
+        positive_ranks_one_indexed <= precision_rank) / precision_rank
 
   return precisions, recalls
 
@@ -272,8 +274,8 @@ def ComputeMetrics(sorted_index_ids, ground_truth, desired_pr_ranks):
 
   if sorted_desired_pr_ranks[-1] > num_index_images:
     raise ValueError(
-      'Requested PR ranks up to %d, however there are only %d images' %
-      (sorted_desired_pr_ranks[-1], num_index_images))
+        'Requested PR ranks up to %d, however there are only %d images' %
+        (sorted_desired_pr_ranks[-1], num_index_images))
 
   # Instantiate all outputs, then loop over each query and gather metrics.
   mean_average_precision = 0.0
@@ -295,7 +297,7 @@ def ComputeMetrics(sorted_index_ids, ground_truth, desired_pr_ranks):
       continue
 
     positive_ranks = np.arange(num_index_images)[np.in1d(
-      sorted_index_ids[i], ok_index_images)]
+        sorted_index_ids[i], ok_index_images)]
     junk_ranks = np.arange(num_index_images)[np.in1d(sorted_index_ids[i],
                                                      junk_index_images)]
 
@@ -335,9 +337,9 @@ def SaveMetricsFile(mean_average_precision, mean_precisions, mean_recalls,
   with tf.io.gfile.GFile(output_path, 'w') as f:
     for k in sorted(mean_average_precision.keys()):
       f.write('{}\n  mAP={}\n  mP@k{} {}\n  mR@k{} {}\n'.format(
-        k, np.around(mean_average_precision[k] * 100, decimals=2),
-        np.array(pr_ranks), np.around(mean_precisions[k] * 100, decimals=2),
-        np.array(pr_ranks), np.around(mean_recalls[k] * 100, decimals=2)))
+          k, np.around(mean_average_precision[k] * 100, decimals=2),
+          np.array(pr_ranks), np.around(mean_precisions[k] * 100, decimals=2),
+          np.array(pr_ranks), np.around(mean_recalls[k] * 100, decimals=2)))
 
 
 def _ParseSpaceSeparatedStringsInBrackets(line, prefixes, ind):
@@ -378,8 +380,8 @@ def _ParsePrRanks(line):
     ValueError: If input line is malformed.
   """
   return [
-    int(pr_rank) for pr_rank in _ParseSpaceSeparatedStringsInBrackets(
-      line, ['  mP@k['], 0) if pr_rank
+      int(pr_rank) for pr_rank in _ParseSpaceSeparatedStringsInBrackets(
+          line, ['  mP@k['], 0) if pr_rank
   ]
 
 
@@ -397,8 +399,8 @@ def _ParsePrScores(line, num_pr_ranks):
     ValueError: If input line is malformed.
   """
   pr_scores = [
-    float(pr_score) for pr_score in _ParseSpaceSeparatedStringsInBrackets(
-      line, ('  mP@k[', '  mR@k['), 1) if pr_score
+      float(pr_score) for pr_score in _ParseSpaceSeparatedStringsInBrackets(
+          line, ('  mP@k[', '  mR@k['), 1) if pr_score
   ]
 
   if len(pr_scores) != num_pr_ranks:
@@ -430,8 +432,8 @@ def ReadMetricsFile(metrics_path):
 
   if len(file_contents_stripped) % 4:
     raise ValueError(
-      'Malformed input %s: number of lines must be a multiple of 4, '
-      'but it is %d' % (metrics_path, len(file_contents_stripped)))
+        'Malformed input %s: number of lines must be a multiple of 4, '
+        'but it is %d' % (metrics_path, len(file_contents_stripped)))
 
   mean_average_precision = {}
   pr_ranks = []
@@ -442,13 +444,13 @@ def ReadMetricsFile(metrics_path):
     protocol = file_contents_stripped[i]
     if protocol in protocols:
       raise ValueError(
-        'Malformed input %s: protocol %s is found a second time' %
-        (metrics_path, protocol))
+          'Malformed input %s: protocol %s is found a second time' %
+          (metrics_path, protocol))
     protocols.add(protocol)
 
     # Parse mAP.
     mean_average_precision[protocol] = float(
-      file_contents_stripped[i + 1].split('=')[1]) / 100.0
+        file_contents_stripped[i + 1].split('=')[1]) / 100.0
 
     # Parse (or check consistency of) pr_ranks.
     parsed_pr_ranks = _ParsePrRanks(file_contents_stripped[i + 2])
@@ -461,18 +463,18 @@ def ReadMetricsFile(metrics_path):
 
     # Parse mean precisions.
     mean_precisions[protocol] = np.array(
-      _ParsePrScores(file_contents_stripped[i + 2], len(pr_ranks)),
-      dtype=float) / 100.0
+        _ParsePrScores(file_contents_stripped[i + 2], len(pr_ranks)),
+        dtype=float) / 100.0
 
     # Parse mean recalls.
     mean_recalls[protocol] = np.array(
-      _ParsePrScores(file_contents_stripped[i + 3], len(pr_ranks)),
-      dtype=float) / 100.0
+        _ParsePrScores(file_contents_stripped[i + 3], len(pr_ranks)),
+        dtype=float) / 100.0
 
   return mean_average_precision, pr_ranks, mean_precisions, mean_recalls
 
 
-def create_config_for_test_dataset(dataset, dir_main):
+def CreateConfigForTestDataset(dataset, dir_main):
   """Creates the configuration dictionary for the test dataset.
 
   Args:
@@ -482,8 +484,8 @@ def create_config_for_test_dataset(dataset, dir_main):
   Returns:
     cfg: Dataset configuration in a form of dictionary. The configuration
       includes:
-      `gnd_fname` - path to the ground truth file for teh dataset,
-      `ext` and `qext` - image extentions for the images in the test dataset
+      `gnd_fname` - path to the ground truth file for the dataset,
+      `ext` and `qext` - image extensions for the images in the test dataset
       and the query images,
       `dir_data` - path to the folder containing ground truth files,
       `dir_images` - path to the folder containing images,
@@ -496,16 +498,15 @@ def create_config_for_test_dataset(dataset, dir_main):
   Raises:
     ValueError: If an unknown dataset name is provided as an argument.
   """
-  DATASETS = ['roxford5k', 'rparis6k']
   dataset = dataset.lower()
 
-  def _config_imname(cfg, i):
+  def _ConfigImname(cfg, i):
     return os.path.join(cfg['dir_images'], cfg['imlist'][i] + cfg['ext'])
 
-  def _config_qimname(cfg, i):
+  def _ConfigQimname(cfg, i):
     return os.path.join(cfg['dir_images'], cfg['qimlist'][i] + cfg['qext'])
 
-  if dataset not in DATASETS:
+  if dataset not in DATASET_NAMES:
     raise ValueError('Unknown dataset: {}!'.format(dataset))
 
   # Loading imlist, qimlist, and gnd in configuration as a dictionary.
@@ -526,8 +527,8 @@ def create_config_for_test_dataset(dataset, dir_main):
   cfg['n'] = len(cfg['imlist'])
   cfg['nq'] = len(cfg['qimlist'])
 
-  cfg['im_fname'] = _config_imname
-  cfg['qim_fname'] = _config_qimname
+  cfg['im_fname'] = _ConfigImname
+  cfg['qim_fname'] = _ConfigQimname
 
   cfg['dataset'] = dataset
 

research/delf/delf/python/datasets/revisited_op/dataset_test.py

@@ -24,7 +24,7 @@ from absl import flags
 import numpy as np
 import tensorflow as tf
 
-from delf.python.detect_to_retrieve import dataset
+from delf.python.datasets.revisited_op import dataset
 
 FLAGS = flags.FLAGS
 

research/delf/delf/python/datasets/utils.py

@@ -19,7 +19,7 @@ import numpy as np
 from PIL import Image
 
 import tensorflow as tf
-from delf.python import utils as image_loading_utils
+from delf import utils as image_loading_utils
 
 
 def pil_imagenet_loader(path, imsize, bounding_box=None, preprocess=True):
@@ -32,7 +32,7 @@ def pil_imagenet_loader(path, imsize, bounding_box=None, preprocess=True):
     preprocess: Bool, whether to preprocess the images in respect to the
       ImageNet dataset.
 
-  Returns: 
+  Returns:
     image: `Tensor`, image in ImageNet suitable format.
   """
   img = image_loading_utils.RgbLoader(path)

research/delf/delf/python/datasets/utils_test.py

@@ -43,8 +43,8 @@ class UtilsTest(tf.test.TestCase):
 
     max_img_size = 1024
     # Load the saved dummy image.
-    img = image_loading_utils.default_loader(filename, imsize=max_img_size,
-                                             preprocess=False)
+    img = image_loading_utils.default_loader(
+        filename, imsize=max_img_size, preprocess=False)
 
     # Make sure the values are the same before and after loading.
     self.assertAllEqual(np.array(img_out), img)
@@ -63,9 +63,10 @@ class UtilsTest(tf.test.TestCase):
     # Load the saved dummy image.
     expected_size = 400
     img = image_loading_utils.default_loader(
-      filename, imsize=max_img_size,
-      bounding_box=[120, 120, 120 + expected_size, 120 + expected_size],
-      preprocess=False)
+        filename,
+        imsize=max_img_size,
+        bounding_box=[120, 120, 120 + expected_size, 120 + expected_size],
+        preprocess=False)
 
     # Check that the final shape is as expected.
     self.assertAllEqual(tf.shape(img), [expected_size, expected_size, 3])

research/delf/delf/python/delg/extract_features.py

@@ -41,7 +41,7 @@ from delf import delf_config_pb2
 from delf import datum_io
 from delf import feature_io
 from delf import utils
-from delf.python.detect_to_retrieve import dataset
+from delf.python.datasets.revisited_op import dataset
 from delf import extractor
 
 FLAGS = flags.FLAGS

research/delf/delf/python/delg/perform_retrieval.py

@@ -28,7 +28,7 @@ import numpy as np
 import tensorflow as tf
 
 from delf import datum_io
-from delf.python.detect_to_retrieve import dataset
+from delf.python.datasets.revisited_op import dataset
 from delf.python.detect_to_retrieve import image_reranking
 
 FLAGS = flags.FLAGS

research/delf/delf/python/detect_to_retrieve/cluster_delf_features.py

@@ -34,12 +34,12 @@ import os
 import sys
 import time
 
+from absl import app
 import numpy as np
 import tensorflow as tf
 
-from tensorflow.python.platform import app
 from delf import feature_io
-from delf.python.detect_to_retrieve import dataset
+from delf.python.datasets.revisited_op import dataset
 
 cmd_args = None
 

research/delf/delf/python/detect_to_retrieve/extract_aggregation.py

@@ -25,9 +25,9 @@ from __future__ import print_function
 import argparse
 import sys
 
-from tensorflow.python.platform import app
+from absl import app
+from delf.python.datasets.revisited_op import dataset
 from delf.python.detect_to_retrieve import aggregation_extraction
-from delf.python.detect_to_retrieve import dataset
 
 cmd_args = None
 

research/delf/delf/python/detect_to_retrieve/extract_index_boxes_and_features.py

@@ -31,9 +31,9 @@ import argparse
 import os
 import sys
 
-from tensorflow.python.platform import app
+from absl import app
+from delf.python.datasets.revisited_op import dataset
 from delf.python.detect_to_retrieve import boxes_and_features_extraction
-from delf.python.detect_to_retrieve import dataset
 
 cmd_args = None
 

research/delf/delf/python/detect_to_retrieve/extract_query_features.py

@@ -31,15 +31,15 @@ import os
 import sys
 import time
 
+from absl import app
 import numpy as np
 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_io
 from delf import utils
-from delf.python.detect_to_retrieve import dataset
+from delf.python.datasets.revisited_op import dataset
 from delf import extractor
 
 cmd_args = None
@@ -76,8 +76,8 @@ def main(argv):
     query_image_name = query_list[i]
     input_image_filename = os.path.join(cmd_args.images_dir,
                                         query_image_name + _IMAGE_EXTENSION)
-    output_feature_filename = os.path.join(
-        cmd_args.output_features_dir, query_image_name + _DELF_EXTENSION)
+    output_feature_filename = os.path.join(cmd_args.output_features_dir,
+                                           query_image_name + _DELF_EXTENSION)
     if tf.io.gfile.exists(output_feature_filename):
       print(f'Skipping {query_image_name}')
       continue
@@ -94,8 +94,7 @@ def main(argv):
     attention_out = extracted_features['local_features']['attention']
 
     feature_io.WriteToFile(output_feature_filename, locations_out,
-                           feature_scales_out, descriptors_out,
-                           attention_out)
+                           feature_scales_out, descriptors_out, attention_out)
 
   elapsed = (time.time() - start)
   print('Processed %d query images in %f seconds' % (num_images, elapsed))

research/delf/delf/python/detect_to_retrieve/perform_retrieval.py

@@ -24,15 +24,15 @@ import os
 import sys
 import time
 
+from absl import app
 import numpy as np
 import tensorflow as tf
 
 from google.protobuf import text_format
-from tensorflow.python.platform import app
 from delf import aggregation_config_pb2
 from delf import datum_io
 from delf import feature_aggregation_similarity
-from delf.python.detect_to_retrieve import dataset
+from delf.python.datasets.revisited_op import dataset
 from delf.python.detect_to_retrieve import image_reranking
 
 cmd_args = None

research/delf/delf/python/examples/extract_boxes.py

@@ -27,12 +27,12 @@ import os
 import sys
 import time
 
+from absl import app
 import matplotlib.patches as patches
 import matplotlib.pyplot as plt
 import numpy as np
 import tensorflow as tf
 
-from tensorflow.python.platform import app
 from delf import box_io
 from delf import utils
 from delf import detector
@@ -153,17 +153,14 @@ def main(argv):
       print('Starting to detect objects in images...')
     elif i % _STATUS_CHECK_ITERATIONS == 0:
       elapsed = (time.time() - start)
-      print(
-          f'Processing image {i} out of {num_images}, last '
-          f'{_STATUS_CHECK_ITERATIONS} images took {elapsed} seconds'
-          )
+      print(f'Processing image {i} out of {num_images}, last '
+            f'{_STATUS_CHECK_ITERATIONS} images took {elapsed} seconds')
       start = time.time()
 
     # If descriptor already exists, skip its computation.
     base_boxes_filename, _ = os.path.splitext(os.path.basename(image_path))
     out_boxes_filename = base_boxes_filename + _BOX_EXT
-    out_boxes_fullpath = os.path.join(cmd_args.output_dir,
-                                      out_boxes_filename)
+    out_boxes_fullpath = os.path.join(cmd_args.output_dir, out_boxes_filename)
     if tf.io.gfile.exists(out_boxes_fullpath):
       print(f'Skipping {image_path}')
       continue
@@ -173,8 +170,7 @@ def main(argv):
     # Extract and save boxes.
     (boxes_out, scores_out, class_indices_out) = detector_fn(im)
     (selected_boxes, selected_scores,
-     selected_class_indices) = _FilterBoxesByScore(boxes_out[0],
-                                                   scores_out[0],
+     selected_class_indices) = _FilterBoxesByScore(boxes_out[0], scores_out[0],
                                                    class_indices_out[0],
                                                    cmd_args.detector_thresh)
 
@@ -182,8 +178,7 @@ def main(argv):
                        selected_class_indices)
     if cmd_args.output_viz_dir:
       out_viz_filename = base_boxes_filename + _VIZ_SUFFIX
-      out_viz_fullpath = os.path.join(cmd_args.output_viz_dir,
-                                      out_viz_filename)
+      out_viz_fullpath = os.path.join(cmd_args.output_viz_dir, out_viz_filename)
       _PlotBoxesAndSaveImage(im[0], selected_boxes, out_viz_fullpath)
 
 

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

@@ -27,12 +27,12 @@ import os
 import sys
 import time
 
+from absl import app
 import numpy as np
 from six.moves import range
 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_io
 from delf import utils
@@ -87,10 +87,8 @@ def main(unused_argv):
       print('Starting to extract DELF features from images...')
     elif i % _STATUS_CHECK_ITERATIONS == 0:
       elapsed = (time.time() - start)
-      print(
-          f'Processing image {i} out of {num_images}, last '
-          f'{_STATUS_CHECK_ITERATIONS} images took {elapsed} seconds'
-      )
+      print(f'Processing image {i} out of {num_images}, last '
+            f'{_STATUS_CHECK_ITERATIONS} images took {elapsed} seconds')
       start = time.time()
 
     # If descriptor already exists, skip its computation.

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

@@ -28,6 +28,7 @@ from __future__ import print_function
 import argparse
 import sys
 
+from absl import app
 import matplotlib
 # Needed before pyplot import for matplotlib to work properly.
 matplotlib.use('Agg')
@@ -39,7 +40,6 @@ from skimage import feature
 from skimage import measure
 from skimage import transform
 
-from tensorflow.python.platform import app
 from delf import feature_io
 
 cmd_args = None

research/delf/delf/python/pooling_layers/pooling.py

@@ -31,6 +31,7 @@ class MAC(tf.keras.layers.Layer):
     Args:
       x: [B, H, W, D] A float32 Tensor.
       axis: Dimensions to reduce. By default, dimensions [1, 2] are reduced.
+
     Returns:
       output: [B, D] A float32 Tensor.
     """
@@ -99,26 +100,30 @@ class GeM(tf.keras.layers.Layer):
 
 
 class GeMPooling2D(tf.keras.layers.Layer):
+  """Generalized mean pooling (GeM) pooling operation for spatial data."""
 
-  def __init__(self, power=20., pool_size=(2, 2), strides=None,
-               padding='valid', data_format='channels_last'):
-    """Generalized mean pooling (GeM) pooling operation for spatial data.
+  def __init__(self,
+               power=20.,
+               pool_size=(2, 2),
+               strides=None,
+               padding='valid',
+               data_format='channels_last'):
+    """Initialization of GeMPooling2D.
 
     Args:
       power: Float, power > 0. is an inverse exponent parameter (GeM power).
-      pool_size: Integer or tuple of 2 integers, factors by which to
-        downscale (vertical, horizontal)
-      strides: Integer, tuple of 2 integers, or None. Strides values.
-        If None, it will default to `pool_size`.
-      padding: One of `valid` or `same`. `valid` means no padding.
-        `same` results in padding evenly to the left/right or up/down of
-        the input such that output has the same height/width dimension as the
-        input.
+      pool_size: Integer or tuple of 2 integers, factors by which to downscale
+        (vertical, horizontal)
+      strides: Integer, tuple of 2 integers, or None. Strides values. If None,
+        it will default to `pool_size`.
+      padding: One of `valid` or `same`. `valid` means no padding. `same`
+        results in padding evenly to the left/right or up/down of the input such
+        that output has the same height/width dimension as the input.
       data_format: A string, one of `channels_last` (default) or
         `channels_first`. The ordering of the dimensions in the inputs.
-        `channels_last` corresponds to inputs with shape `(batch, height,
-        width, channels)` while `channels_first` corresponds to inputs with
-        shape `(batch, channels, height, width)`.
+        `channels_last` corresponds to inputs with shape `(batch, height, width,
+        channels)` while `channels_first` corresponds to inputs with shape
+        `(batch, channels, height, width)`.
     """
     super(GeMPooling2D, self).__init__()
     self.power = power
@@ -126,15 +131,16 @@ class GeMPooling2D(tf.keras.layers.Layer):
     self.pool_size = pool_size
     self.strides = strides
     self.padding = padding.upper()
-    data_format_conv = {'channels_last': 'NHWC',
-                        'channels_first': 'NCHW',
-                        }
+    data_format_conv = {
+        'channels_last': 'NHWC',
+        'channels_first': 'NCHW',
+    }
     self.data_format = data_format_conv[data_format]
 
   def call(self, x):
     tmp = tf.pow(x, self.power)
-    tmp = tf.nn.avg_pool(tmp, self.pool_size, self.strides,
-                         self.padding, self.data_format)
+    tmp = tf.nn.avg_pool(tmp, self.pool_size, self.strides, self.padding,
+                         self.data_format)
     out = tf.pow(tmp, 1. / self.power)
     return out
 

research/delf/delf/python/training/global_features_utils.py

@@ -19,7 +19,6 @@ import os
 from absl import logging
 
 import numpy as np
-from tensorboard import program
 import tensorflow as tf
 
 from delf.python.datasets.revisited_op import dataset
@@ -52,8 +51,11 @@ class AverageMeter():
     self.avg = self.sum / self.count
 
 
-def compute_metrics_and_print(dataset_name, sorted_index_ids, ground_truth,
-                              desired_pr_ranks=None, log=True):
+def compute_metrics_and_print(dataset_name,
+                              sorted_index_ids,
+                              ground_truth,
+                              desired_pr_ranks=None,
+                              log=True):
   """Computes and logs ground-truth metrics for Revisited datasets.
 
   Args:
@@ -68,6 +70,7 @@ def compute_metrics_and_print(dataset_name, sorted_index_ids, ground_truth,
       ranks to be reported. E.g., if precision@1/recall@1 and
       precision@10/recall@10 are desired, this should be set to [1, 10]. The
       largest item should be <= #sorted_index_ids. Default: [1, 5, 10].
+    log: Whether to log results using logging.info().
 
   Returns:
     mAP: (metricsE, metricsM, metricsH) Tuple of the metrics for different
@@ -82,8 +85,7 @@ def compute_metrics_and_print(dataset_name, sorted_index_ids, ground_truth,
   Raises:
     ValueError: If an unknown dataset name is provided as an argument.
   """
-  _DATASETS = ['roxford5k', 'rparis6k']
-  if dataset not in _DATASETS:
+  if dataset not in dataset.DATASET_NAMES:
     raise ValueError('Unknown dataset: {}!'.format(dataset))
 
   if desired_pr_ranks is None:
@@ -94,24 +96,25 @@ def compute_metrics_and_print(dataset_name, sorted_index_ids, ground_truth,
 
   metrics_easy = dataset.ComputeMetrics(sorted_index_ids, easy_ground_truth,
                                         desired_pr_ranks)
-  metrics_medium = dataset.ComputeMetrics(sorted_index_ids,
-                                          medium_ground_truth,
+  metrics_medium = dataset.ComputeMetrics(sorted_index_ids, medium_ground_truth,
                                           desired_pr_ranks)
   metrics_hard = dataset.ComputeMetrics(sorted_index_ids, hard_ground_truth,
                                         desired_pr_ranks)
 
   debug_and_log(
-    '>> {}: mAP E: {}, M: {}, H: {}'.format(
-      dataset_name, np.around(metrics_easy[0] * 100, decimals=2),
-      np.around(metrics_medium[0] * 100, decimals=2),
-      np.around(metrics_hard[0] * 100, decimals=2)), log=log)
+      '>> {}: mAP E: {}, M: {}, H: {}'.format(
+          dataset_name, np.around(metrics_easy[0] * 100, decimals=2),
+          np.around(metrics_medium[0] * 100, decimals=2),
+          np.around(metrics_hard[0] * 100, decimals=2)),
+      log=log)
 
   debug_and_log(
-    '>> {}: mP@k{} E: {}, M: {}, H: {}'.format(
-      dataset_name, desired_pr_ranks,
-      np.around(metrics_easy[1] * 100, decimals=2),
-      np.around(metrics_medium[1] * 100, decimals=2),
-      np.around(metrics_hard[1] * 100, decimals=2)), log=log)
+      '>> {}: mP@k{} E: {}, M: {}, H: {}'.format(
+          dataset_name, desired_pr_ranks,
+          np.around(metrics_easy[1] * 100, decimals=2),
+          np.around(metrics_medium[1] * 100, decimals=2),
+          np.around(metrics_hard[1] * 100, decimals=2)),
+      log=log)
 
   return metrics_easy, metrics_medium, metrics_hard
 
@@ -151,8 +154,8 @@ def debug_and_log(msg, debug=True, log=True, debug_on_the_same_line=False):
     msg: String, message to be logged.
     debug: Bool, if True, will print `msg` to stdout.
     log: Bool, if True, will redirect `msg` to the logfile.
-    debug_on_the_same_line: Bool, if True, will print `msg` to stdout without
-      a new line. When using this mode, logging to a logfile is disabled.
+    debug_on_the_same_line: Bool, if True, will print `msg` to stdout without a
+      new line. When using this mode, logging to a logfile is disabled.
   """
   if debug_on_the_same_line:
     print(msg, end='')
@@ -163,34 +166,23 @@ def debug_and_log(msg, debug=True, log=True, debug_on_the_same_line=False):
     logging.info(msg)
 
 
-def launch_tensorboard(log_dir):
-  """Runs tensorboard with the given `log_dir`.
-  
-  Args:
-    log_dir: String, directory to start tensorboard in.
-  """
-  tb = program.TensorBoard()
-  tb.configure(argv=[None, '--logdir', log_dir])
-  url = tb.launch()
-  debug_and_log("Launching Tensorboard: {}".format(url))
-
-
 def get_standard_keras_models():
   """Gets the standard keras model names.
 
   Returns:
     model_names: List, names of the standard keras models.
   """
-  model_names = sorted(name for name in tf.keras.applications.__dict__
-                       if not name.startswith("__")
-                       and callable(tf.keras.applications.__dict__[name]))
+  model_names = sorted(
+      name for name in tf.keras.applications.__dict__
+      if not name.startswith('__') and
+      callable(tf.keras.applications.__dict__[name]))
   return model_names
 
 
-def create_model_directory(training_dataset, arch, pool, whitening,
-                           pretrained, loss, loss_margin, optimizer, lr,
-                           weight_decay, neg_num, query_size, pool_size,
-                           batch_size, update_every, image_size, directory):
+def create_model_directory(training_dataset, arch, pool, whitening, pretrained,
+                           loss, loss_margin, optimizer, lr, weight_decay,
+                           neg_num, query_size, pool_size, batch_size,
+                           update_every, image_size, directory):
   """Based on the model parameters, creates the model directory.
 
   If the model directory does not exist, the directory is created.
@@ -224,13 +216,14 @@ def create_model_directory(training_dataset, arch, pool, whitening,
   if not pretrained:
     folder += '_notpretrained'
   folder += ('_{}_m{:.2f}_{}_lr{:.1e}_wd{:.1e}_nnum{}_qsize{}_psize{}_bsize{}'
-             '_uevery{}_imsize{}').format(
-            loss, loss_margin, optimizer, lr, weight_decay, neg_num,
-            query_size, pool_size, batch_size, update_every, image_size)
+             '_uevery{}_imsize{}').format(loss, loss_margin, optimizer, lr,
+                                          weight_decay, neg_num, query_size,
+                                          pool_size, batch_size, update_every,
+                                          image_size)
 
   folder = os.path.join(directory, folder)
   debug_and_log(
-    '>> Creating directory if does not exist:\n>> \'{}\''.format(folder))
+      '>> Creating directory if does not exist:\n>> \'{}\''.format(folder))
   if not os.path.exists(folder):
     os.makedirs(folder)
   return folder