Dataset utilities added. (#9796)

Author: jejarosl

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

@@ -0,0 +1,71 @@
+# Lint as: python3
+# Copyright 2021 The TensorFlow Authors All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+# ==============================================================================
+"""Supporting functions for data loading."""
+
+import numpy as np
+from PIL import Image
+
+import tensorflow as tf
+from delf.python import utils as image_loading_utils
+
+
+def pil_imagenet_loader(path, imsize, bounding_box=None, normalize=True):
+  """Pillow loader for the images.
+
+  Args:
+    path: Path to image to be loaded.
+    imsize: Integer, defines the maximum size of longer image side.
+    bounding_box: (x1,y1,x2,y2) tuple to crop the query image.
+    normalize: Bool, whether to normalize the image.
+
+  Returns: 
+    image: `Tensor`, image in ImageNet suitable format.
+  """
+  img = image_loading_utils.RgbLoader(path)
+
+  if bounding_box is not None:
+    imfullsize = max(img.size)
+    img = img.crop(bounding_box)
+    imsize = imsize * max(img.size) / imfullsize
+
+  # Unlike `resize`, `thumbnail` resizes to the largest size that preserves
+  # the aspect ratio, making sure that the output image does not exceed the
+  # original image size and the size specified in the arguments of thumbnail.
+  img.thumbnail((imsize, imsize), Image.ANTIALIAS)
+  img = np.array(img)
+
+  if normalize:
+    # Preprocessing for ImageNet data. Converts the images from RGB to BGR,
+    # then zero-centers each color channel with respect to the ImageNet
+    # dataset, without scaling.
+    tf.keras.applications.imagenet_utils.preprocess_input(img, mode='caffe')
+
+  return img
+
+
+def default_loader(path, imsize, bounding_box=None, normalize=True):
+  """Default loader for the images is using Pillow.
+
+  Args:
+    path: Path to image to be loaded.
+    imsize: Integer, defines the maximum size of longer image side.
+    bounding_box: (x1,y1,x2,y2) tuple to crop the query image.
+
+  Returns:
+    image: `Tensor`, image in ImageNet suitable format.
+  """
+  img = pil_imagenet_loader(path, imsize, bounding_box, normalize)
+  return img

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

@@ -0,0 +1,76 @@
+# Lint as: python3
+# Copyright 2021 The TensorFlow Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+# ==============================================================================
+"""Tests for dataset utilities."""
+
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+
+import os
+
+from absl import flags
+import numpy as np
+from PIL import Image
+import tensorflow as tf
+
+from delf.python.datasets import utils as image_loading_utils
+
+FLAGS = flags.FLAGS
+
+
+class UtilsTest(tf.test.TestCase):
+
+  def testDefaultLoader(self):
+    # Create a dummy image.
+    dummy_image = np.random.rand(1024, 750, 3) * 255
+    img_out = Image.fromarray(dummy_image.astype('uint8')).convert('RGB')
+    filename = os.path.join(FLAGS.test_tmpdir, 'test_image.png')
+    # Save the dummy image.
+    img_out.save(filename)
+
+    max_img_size = 1024
+    # Load the saved dummy image.
+    img = image_loading_utils.default_loader(filename, imsize=max_img_size,
+                                     normalize=False)
+
+    # Make sure the values are the same before and after loading.
+    self.assertAllEqual(np.array(img_out), img)
+
+    self.assertAllLessEqual(tf.shape(img), max_img_size)
+
+  def testDefaultLoaderWithBoundingBox(self):
+    # Create a dummy image.
+    dummy_image = np.random.rand(1024, 750, 3) * 255
+    img_out = Image.fromarray(dummy_image.astype('uint8')).convert('RGB')
+    filename = os.path.join(FLAGS.test_tmpdir, 'test_image.png')
+    # Save the dummy image.
+    img_out.save(filename)
+
+    max_img_size = 1024
+    # 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],
+                                     normalize=False)
+
+    # Check the shape.
+    self.assertAllEqual(tf.shape(img), [expected_size, expected_size, 3])
+
+
+if __name__ == '__main__':
+  tf.test.main()