From 3de425d51782cb4395d634ed14e51efed2f4e28b Mon Sep 17 00:00:00 2001
From: Hermann <blumh@ethz.ch>
Date: Tue, 27 Aug 2019 10:58:31 +0200
Subject: [PATCH 01/11] sort imports

---
 tensorflow_datasets/image/__init__.py | 1 +
 1 file changed, 1 insertion(+)

diff --git a/tensorflow_datasets/image/__init__.py b/tensorflow_datasets/image/__init__.py
index 625c2b3f495..f3344f86841 100644
--- a/tensorflow_datasets/image/__init__.py
+++ b/tensorflow_datasets/image/__init__.py
@@ -53,6 +53,7 @@
 from tensorflow_datasets.image.imagenet2012_corrupted import Imagenet2012Corrupted
 from tensorflow_datasets.image.kitti import Kitti
 from tensorflow_datasets.image.lfw import LFW
+from tensorflow_datasets.image.lost_and_found import LostAndFound
 from tensorflow_datasets.image.lsun import Lsun
 from tensorflow_datasets.image.malaria import Malaria
 from tensorflow_datasets.image.mnist import EMNIST

From 818744cb23505f02898e785f20a045b9e2964ef8 Mon Sep 17 00:00:00 2001
From: Hermann <blumh@ethz.ch>
Date: Tue, 27 Aug 2019 10:55:21 +0200
Subject: [PATCH 02/11] ititial working version of lostandfound

---
 tensorflow_datasets/image/__init__.py         |   1 +
 tensorflow_datasets/image/lost_and_found.py   | 228 ++++++++++++++++++
 .../image/lost_and_found_test.py              |  60 +++++
 tensorflow_datasets/testing/cityscapes.py     |  59 +++++
 .../url_checksums/lost_and_found.txt          |   6 +
 5 files changed, 354 insertions(+)
 create mode 100644 tensorflow_datasets/image/lost_and_found.py
 create mode 100644 tensorflow_datasets/image/lost_and_found_test.py
 create mode 100644 tensorflow_datasets/testing/cityscapes.py
 create mode 100644 tensorflow_datasets/url_checksums/lost_and_found.txt

diff --git a/tensorflow_datasets/image/__init__.py b/tensorflow_datasets/image/__init__.py
index f3344f86841..ed9c567f9f2 100644
--- a/tensorflow_datasets/image/__init__.py
+++ b/tensorflow_datasets/image/__init__.py
@@ -83,3 +83,4 @@
 from tensorflow_datasets.image.uc_merced import UcMerced
 from tensorflow_datasets.image.visual_domain_decathlon import VisualDomainDecathlon
 from tensorflow_datasets.image.voc import Voc2007
+from tensorflow_datasets.image.lost_and_found import LostAndFound  # TODO(lost_and_found) Sort alphabetically
diff --git a/tensorflow_datasets/image/lost_and_found.py b/tensorflow_datasets/image/lost_and_found.py
new file mode 100644
index 00000000000..725a7c5c11e
--- /dev/null
+++ b/tensorflow_datasets/image/lost_and_found.py
@@ -0,0 +1,228 @@
+"""TODO(lost_and_found): Add a description here."""
+
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+
+from os import path, listdir
+import re
+
+import tensorflow as tf
+import tensorflow_datasets as tfds
+from tensorflow_datasets.core import api_utils
+
+
+_CITATION = """
+@inproceedings{pinggera2016lost,
+  title={Lost and found: detecting small road hazards for self-driving vehicles},
+  author={Pinggera, Peter and Ramos, Sebastian and Gehrig, Stefan and Franke, Uwe and Rother, Carsten and Mester, Rudolf},
+  booktitle={2016 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)},
+  year={2016}
+}
+"""
+
+_DESCRIPTION = """
+The LostAndFound Dataset addresses the problem of detecting unexpected small obstacles on
+the road often caused by lost cargo. The dataset comprises 112 stereo video sequences
+with 2104 annotated frames (picking roughly every tenth frame from the recorded data).
+
+The dataset is designed analogous to the 'Cityscapes' dataset. The datset provides:
+- stereo image pairs in either 8 or 16 bit color resolution
+- precomputed disparity maps
+- coarse semantic labels for objects and street
+"""
+
+
+class LostAndFoundConfig(tfds.core.BuilderConfig):
+  '''BuilderConfig for 'Lost and Found'
+
+    Args:
+      right_images (bool): Enables right images for stereo image tasks.
+      segmentation_labels (bool): Enables image segmentation labels.
+      disparity_maps (bool): Enables disparity maps.
+      use_16bit (bool): Loads 16 bit (rgb) images instead of 8bit.
+  '''
+
+  @api_utils.disallow_positional_args
+  def __init__(self, right_images=False, segmentation_labels=False, instance_ids=False,
+               disparity_maps=False, use_16bit=False, **kwargs):
+    super().__init__(**kwargs)
+
+    self.ignored_ids = set()
+
+    self.features = ['image_left']
+    if right_images:
+      self.features.append('image_right')
+      # this image causes IO errors
+      # self.ignored_ids.add('06_Galgenbergstr_40_000000_000040')
+    if segmentation_labels:
+      self.features.append('segmentation_label')
+    if disparity_maps:
+      self.features.append('disparity_map')
+
+    self.left_image_string = 'leftImg{}bit'.format('16' if use_16bit else '8')
+    self.right_image_string = 'rightImg{}bit'.format('16' if use_16bit else '8')
+
+
+class LostAndFound(tfds.core.GeneratorBasedBuilder):
+  """TODO(lost_and_found): Short description of my dataset."""
+
+  VERSION = tfds.core.Version('1.0.0')
+
+  BUILDER_CONFIGS = [
+      LostAndFoundConfig(
+          name='semantic_segmentation',
+          description='Lost and Found semantic segmentation dataset.',
+          version="1.0.0",
+          right_images=False,
+          segmentation_labels=True,
+          instance_ids=False,
+          disparity_maps=False,
+          use_16bit=False,
+      ),
+      LostAndFoundConfig(
+          name='stereo_disparity',
+          description='Lost and Found stereo images and disparity maps.',
+          version="1.0.0",
+          right_images=True,
+          segmentation_labels=False,
+          instance_ids=False,
+          disparity_maps=True,
+          use_16bit=False,
+      ),
+      LostAndFoundConfig(
+          name='full',
+          description='Full Lost and Found dataset.',
+          version="1.0.0",
+          right_images=True,
+          segmentation_labels=True,
+          instance_ids=True,
+          disparity_maps=True,
+          use_16bit=False,
+      ),
+      LostAndFoundConfig(
+          name='full_16bit',
+          description='Full Lost and Found dataset.',
+          version="1.0.0",
+          right_images=True,
+          segmentation_labels=True,
+          instance_ids=True,
+          disparity_maps=True,
+          use_16bit=True,
+      )]
+
+  def _info(self):
+    possible_features = {
+        'image_left': tfds.features.Image(shape=(1024, 2048, 3), encoding_format='png'),
+        'image_right': tfds.features.Image(shape=(1024, 2048, 3), encoding_format='png'),
+        'segmentation_label': tfds.features.Image(shape=(1024, 2048, 1),
+                                                  encoding_format='png'),
+        'disparity_map': tfds.features.Image(shape=(1024, 2048, 1),
+                                             encoding_format='png')}
+    return tfds.core.DatasetInfo(
+        builder=self,
+        # This is the description that will appear on the datasets page.
+        description=_DESCRIPTION,
+        # tfds.features.FeatureConnectors
+        features=tfds.features.FeaturesDict({
+            'image_id': tfds.features.Text(),
+            **{feat: possible_features[feat] for feat in self.builder_config.features}}),
+        # Homepage of the dataset for documentation
+        urls=['http://www.6d-vision.com/lostandfounddataset'],
+        citation=_CITATION,
+    )
+
+  def _split_generators(self, dl_manager):
+    """Returns SplitGenerators."""
+    # TODO(lost_and_found): Downloads the data and defines the splits
+    # dl_manager is a tfds.download.DownloadManager that can be used to
+    # download and extract URLs
+    base_url = 'http://www.dhbw-stuttgart.de/~sgehrig/lostAndFoundDataset/{}.zip'
+    download_urls = {
+        'image_left': base_url.format(self.builder_config.left_image_string)}
+    if 'image_right' in self.builder_config.features:
+      download_urls['image_right'] = base_url.format(
+          self.builder_config.right_image_string)
+    if 'segmentation_label' in self.builder_config.features \
+        or 'instance_id' in self.builder_config.features:
+      download_urls['gt'] = base_url.format('gtCoarse')
+    if 'disparity_map' in self.builder_config.features:
+      download_urls['disparity_map'] = base_url.format('disparity')
+    # split into two steps to save space for testing data
+    dl_paths = dl_manager.download(download_urls)
+    dl_paths = dl_manager.extract(dl_paths)
+
+    # point segmentation label and instance IDs both to directory for ground-truth
+    if 'gt' in dl_paths:
+      dl_paths['segmentation_label'] = dl_paths['gt']
+      dl_paths['instance_id'] = dl_paths['gt']
+
+    # first directory in the zipfile, dependent on feature to load
+    sub_dirs = {
+        'image_left': self.builder_config.left_image_string,
+        'image_right': self.builder_config.right_image_string,
+        'segmentation_label': 'gtCoarse',
+        'instance_id': 'gtCoarse',
+        'disparity_map': 'disparity'}
+
+    return [
+        tfds.core.SplitGenerator(
+            name=tfds.Split.TRAIN,
+            # These kwargs will be passed to _generate_examples
+            gen_kwargs={feat: path.join(dl_paths[feat], sub_dirs[feat], 'train')
+                        for feat in self.builder_config.features},
+        ),
+        tfds.core.SplitGenerator(
+            name=tfds.Split.TEST,
+            # These kwargs will be passed to _generate_examples
+            gen_kwargs={feat: path.join(dl_paths[feat], sub_dirs[feat], 'test')
+                        for feat in self.builder_config.features},
+        )
+    ]
+
+  def _generate_examples(self, **paths):
+    """Yields examples."""
+    # different file-suffixes dependent on the feature to load
+    file_suffix = {
+        'image_left': self.builder_config.left_image_string,
+        'image_right': self.builder_config.right_image_string,
+        'segmentation_label': 'gtCoarse_labelIds',
+        'instance_id': 'gtCoarse_instanceIds',
+        'disparity_map': 'disparity'}
+
+    print(paths['image_left'])
+
+    for scene_id in tf.io.gfile.listdir(paths['image_left']):
+      paths_city_root = {feat: path.join(feat_dir, scene_id)
+                         for feat, feat_dir in paths.items()}
+
+      left_city_root = paths_city_root['image_left']
+      for left_img in tf.io.gfile.listdir(left_city_root):
+        image_id = _get_id_from_left_image(left_img)
+
+        if image_id in self.builder_config.ignored_ids:
+          continue
+
+        print(listdir(paths_city_root['segmentation_label']))
+
+        features = {
+            'image_id': image_id,
+            **{feat: path.join(paths_city_root[feat],
+                               '{}_{}.png'.format(image_id, file_suffix[feat]))
+               for feat in paths}}
+
+        yield image_id, features
+
+# Helper functions
+
+
+LEFT_IMAGE_FILE_RE = re.compile(r'(.+)_leftImg(?:8|16)bit\.png')
+
+
+def _get_id_from_left_image(left_image):
+  '''Returns the id of an image file. Used to associate an image file
+  with its corresponding label.
+  Example:
+    'bonn_000001_000019_leftImg8bit' -> 'bonn_000001_000019'
+  '''
+  return LEFT_IMAGE_FILE_RE.match(left_image).group(1)
diff --git a/tensorflow_datasets/image/lost_and_found_test.py b/tensorflow_datasets/image/lost_and_found_test.py
new file mode 100644
index 00000000000..cf9fd9b1423
--- /dev/null
+++ b/tensorflow_datasets/image/lost_and_found_test.py
@@ -0,0 +1,60 @@
+"""TODO(lost_and_found): Add a description here."""
+
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+
+from os import path, listdir, remove
+import tensorflow as tf
+
+from tensorflow_datasets import testing
+from tensorflow_datasets.image import lost_and_found
+from tensorflow_datasets.testing.cityscapes import generate_ids, create_zipfile
+
+
+class LostAndFoundTest(testing.DatasetBuilderTestCase):
+  DATASET_CLASS = lost_and_found.LostAndFound
+  BUILDER_CONFIG_NAMES_TO_TEST = ['semantic_segmentation', 'full']
+  SPLITS = {
+      "train": 4,  # Number of fake train example
+      "test": 2,  # Number of fake test example
+  }
+
+  # If you are calling `download/download_and_extract` with a dict, like:
+  #   dl_manager.download({'some_key': 'http://a.org/out.txt', ...})
+  # then the tests needs to provide the fake output paths relative to the
+  # fake data directory
+  DL_EXTRACT_RESULT = {
+      'image_left': 'leftImg8bit.zip',
+      'image_right': 'rightImg8bit.zip',
+      'disparity_map': 'disparity.zip',
+      'gt': 'gtCoarse.zip'}
+
+
+if __name__ == "__main__":
+  tf.compat.v1.enable_eager_execution()
+
+  # create fake files
+  testing.test_utils.remake_dir(
+      'tensorflow_datasets/testing/test_data/fake_examples/lost_and_found')
+  base_path = 'tensorflow_datasets/testing/test_data/fake_examples/lost_and_found/{}.zip'
+  # generate image ids matching between zipfiles
+  train_ids = [*generate_ids('01_Turmstr_17'), *generate_ids('02_Goethe_Str_6')]
+  test_ids = list(generate_ids('03_Schlossallee_1'))
+  splits = {'train': train_ids, 'test': test_ids}
+  with tf.Graph().as_default():
+    create_zipfile(
+        base_path.format('leftImg8bit'), splits_with_ids=splits, suffixes=['leftImg8bit'])
+    create_zipfile(
+        base_path.format('gtCoarse'), splits_with_ids=splits,
+        suffixes=['gtCoarse_instanceIds', 'gtCoarse_labelIds', 'gtCoarse_color'])
+    create_zipfile(
+        base_path.format('rightImg8bit'), splits_with_ids=splits, suffixes=['rightImg8bit'])
+    create_zipfile(
+        base_path.format('disparity'), splits_with_ids=splits, suffixes=['disparity'])
+
+  testing.test_main()
+
+  # remove fake files
+  testing.test_utils.remake_dir(
+      'tensorflow_datasets/testing/test_data/fake_examples/lost_and_found')
diff --git a/tensorflow_datasets/testing/cityscapes.py b/tensorflow_datasets/testing/cityscapes.py
new file mode 100644
index 00000000000..a5446f06f20
--- /dev/null
+++ b/tensorflow_datasets/testing/cityscapes.py
@@ -0,0 +1,59 @@
+# coding=utf-8
+# Copyright 2019 The TensorFlow Datasets Authors.
+#
+# 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.
+
+"""Helper functions to generate fake Cityscapes-like zip archives for testing."""
+
+from tensorflow_datasets.testing.fake_data_utils import get_random_png
+
+import re
+from os import path
+from zipfile import ZipFile
+from random import randint
+
+
+CITY_IN_ID_RE = re.compile(r'(.+)_[0-9]+_[0-9]+')
+
+
+def generate_ids(city, num=2):
+  for _ in range(num):
+    yield '{}_{:06d}_{:06d}'.format(city, randint(0, 999999), randint(0, 999999))
+
+
+def create_zipfile(zip_filepath, splits_with_ids, suffixes=['leftImg8bit'],
+                   maindir=None):
+  """
+  Generates a zipfile with a cityscapes-like file structure and random pngs.
+
+  Args:
+    zip_filepath (str): filepath to the zip archive that will be created
+    splits_with_ids (Dict[str, List[str]]): data-splits like 'train' or 'val' that map to
+        a list of image ids
+    suffixes (List[str]): suffix per modality that should be created e.g. 'leftImg8bit'
+    maindir (str): name of the root directory of the zipfile, defaults to the name of the
+        zipfile
+  """
+  with ZipFile(zip_filepath, 'w') as z:
+    for split, ids in splits_with_ids.items():
+      if maindir is None:
+        maindir = path.basename(zip_filepath).strip('.zip')
+      split = path.join(maindir, split)
+      for img_id in ids:
+        city = CITY_IN_ID_RE.match(img_id).group(1)
+        for suffix in suffixes:
+          if 'Img' in suffix:
+            img = get_random_png(height=1024, width=2048, channels=3)
+          else:
+            img = get_random_png(height=1024, width=2048, channels=1)
+          z.write(img, path.join(split, city, '{}_{}.png'.format(img_id, suffix)))
diff --git a/tensorflow_datasets/url_checksums/lost_and_found.txt b/tensorflow_datasets/url_checksums/lost_and_found.txt
new file mode 100644
index 00000000000..7fb974baaf3
--- /dev/null
+++ b/tensorflow_datasets/url_checksums/lost_and_found.txt
@@ -0,0 +1,6 @@
+http://www.dhbw-stuttgart.de/~sgehrig/lostAndFoundDataset/disparity.zip 1461824611 1e06350d082f3bd686ff889940ad60ed85bfb1e8aa691a547a259c52fa3b60b1
+http://www.dhbw-stuttgart.de/~sgehrig/lostAndFoundDataset/gtCoarse.zip 37756896 53b6d3ab000f08b1fb59d70c1398eecc4d82a7baf4e9cf74fbf60d1858abe9ac
+http://www.dhbw-stuttgart.de/~sgehrig/lostAndFoundDataset/leftImg16bit.zip 18039875634 f3530514163f30ccafc05210b643ea690c4ba17cdb4497d8a7d4f9c324c71da8
+http://www.dhbw-stuttgart.de/~sgehrig/lostAndFoundDataset/leftImg8bit.zip 5802953400 307f66002023ab597d309963b94990f5b9a8e5735ee729c3292647a66e9f2b18
+http://www.dhbw-stuttgart.de/~sgehrig/lostAndFoundDataset/rightImg16bit.zip 17938768019 db492fc9b9e0adf7a662cd589a6c8ea5cd8cf68d08600b2099d0cd7e0c58f6dd
+http://www.dhbw-stuttgart.de/~sgehrig/lostAndFoundDataset/rightImg8bit.zip 5787134165 d5219f49e730a1ce064a9d118227e71cd39681bcc7f8a87ab4061c86cd7dc6fb

From 43d2f999fbe612ab2067606e845e765817b587da Mon Sep 17 00:00:00 2001
From: Hermann <blumh@ethz.ch>
Date: Tue, 27 Aug 2019 11:06:28 +0200
Subject: [PATCH 03/11] refine docstrings, remove all todos

---
 tensorflow_datasets/image/lost_and_found.py   | 12 +++++-----
 .../image/lost_and_found_test.py              |  2 +-
 tensorflow_datasets/testing/cityscapes.py     | 23 +++++++------------
 3 files changed, 15 insertions(+), 22 deletions(-)

diff --git a/tensorflow_datasets/image/lost_and_found.py b/tensorflow_datasets/image/lost_and_found.py
index 725a7c5c11e..38fdca151b9 100644
--- a/tensorflow_datasets/image/lost_and_found.py
+++ b/tensorflow_datasets/image/lost_and_found.py
@@ -1,4 +1,4 @@
-"""TODO(lost_and_found): Add a description here."""
+"""Lost and Found Road Hazard Dataset."""
 
 from __future__ import absolute_import
 from __future__ import division
@@ -30,6 +30,8 @@
 - stereo image pairs in either 8 or 16 bit color resolution
 - precomputed disparity maps
 - coarse semantic labels for objects and street
+
+Descriptions of the labels are given here: http://www.6d-vision.com/laf_table.pdf
 """
 
 
@@ -39,6 +41,7 @@ class LostAndFoundConfig(tfds.core.BuilderConfig):
     Args:
       right_images (bool): Enables right images for stereo image tasks.
       segmentation_labels (bool): Enables image segmentation labels.
+      instance_ids (bool): Enables instance-id labels.
       disparity_maps (bool): Enables disparity maps.
       use_16bit (bool): Loads 16 bit (rgb) images instead of 8bit.
   '''
@@ -65,7 +68,7 @@ def __init__(self, right_images=False, segmentation_labels=False, instance_ids=F
 
 
 class LostAndFound(tfds.core.GeneratorBasedBuilder):
-  """TODO(lost_and_found): Short description of my dataset."""
+  """Lost and Found Road Hazard Dataset."""
 
   VERSION = tfds.core.Version('1.0.0')
 
@@ -134,10 +137,7 @@ def _info(self):
 
   def _split_generators(self, dl_manager):
     """Returns SplitGenerators."""
-    # TODO(lost_and_found): Downloads the data and defines the splits
-    # dl_manager is a tfds.download.DownloadManager that can be used to
-    # download and extract URLs
-    base_url = 'http://www.dhbw-stuttgart.de/~sgehrig/lostAndFoundDataset/{}.zip'
+        base_url = 'http://www.dhbw-stuttgart.de/~sgehrig/lostAndFoundDataset/{}.zip'
     download_urls = {
         'image_left': base_url.format(self.builder_config.left_image_string)}
     if 'image_right' in self.builder_config.features:
diff --git a/tensorflow_datasets/image/lost_and_found_test.py b/tensorflow_datasets/image/lost_and_found_test.py
index cf9fd9b1423..3bb27f8b3bc 100644
--- a/tensorflow_datasets/image/lost_and_found_test.py
+++ b/tensorflow_datasets/image/lost_and_found_test.py
@@ -1,4 +1,4 @@
-"""TODO(lost_and_found): Add a description here."""
+"""Tests for LostAndFound dataset module."""
 
 from __future__ import absolute_import
 from __future__ import division
diff --git a/tensorflow_datasets/testing/cityscapes.py b/tensorflow_datasets/testing/cityscapes.py
index a5446f06f20..f428b42cd5a 100644
--- a/tensorflow_datasets/testing/cityscapes.py
+++ b/tensorflow_datasets/testing/cityscapes.py
@@ -1,18 +1,3 @@
-# coding=utf-8
-# Copyright 2019 The TensorFlow Datasets Authors.
-#
-# 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.
-
 """Helper functions to generate fake Cityscapes-like zip archives for testing."""
 
 from tensorflow_datasets.testing.fake_data_utils import get_random_png
@@ -27,6 +12,14 @@
 
 
 def generate_ids(city, num=2):
+  """ Generates image ids following the format of the cityscapes dataset.
+
+  Args:
+    city (str): The city/scene the ids belong to, used as a prefix to the id.
+    num (int): Number of random ids to generate.
+  Returns:
+    Generator for id strings.
+  """
   for _ in range(num):
     yield '{}_{:06d}_{:06d}'.format(city, randint(0, 999999), randint(0, 999999))
 

From 37a8f28fe1788a64c38d29d825c8c6cc579928a6 Mon Sep 17 00:00:00 2001
From: Hermann <blumh@ethz.ch>
Date: Tue, 27 Aug 2019 13:06:08 +0200
Subject: [PATCH 04/11] fix style errors

---
 tensorflow_datasets/image/lost_and_found.py   | 35 +++++++--------
 .../image/lost_and_found_test.py              | 43 ++++++++++---------
 tensorflow_datasets/testing/cityscapes.py     | 26 +++++------
 3 files changed, 52 insertions(+), 52 deletions(-)

diff --git a/tensorflow_datasets/image/lost_and_found.py b/tensorflow_datasets/image/lost_and_found.py
index 38fdca151b9..4e746bb65b2 100644
--- a/tensorflow_datasets/image/lost_and_found.py
+++ b/tensorflow_datasets/image/lost_and_found.py
@@ -4,7 +4,7 @@
 from __future__ import division
 from __future__ import print_function
 
-from os import path, listdir
+from os import path
 import re
 
 import tensorflow as tf
@@ -47,19 +47,18 @@ class LostAndFoundConfig(tfds.core.BuilderConfig):
   '''
 
   @api_utils.disallow_positional_args
-  def __init__(self, right_images=False, segmentation_labels=False, instance_ids=False,
-               disparity_maps=False, use_16bit=False, **kwargs):
+  def __init__(self, right_images=False, segmentation_labels=False,
+               instance_ids=False, disparity_maps=False, use_16bit=False,
+               **kwargs):
     super().__init__(**kwargs)
 
-    self.ignored_ids = set()
-
     self.features = ['image_left']
     if right_images:
       self.features.append('image_right')
-      # this image causes IO errors
-      # self.ignored_ids.add('06_Galgenbergstr_40_000000_000040')
     if segmentation_labels:
       self.features.append('segmentation_label')
+    if instance_ids:
+      self.features.append('instance_id')
     if disparity_maps:
       self.features.append('disparity_map')
 
@@ -116,10 +115,14 @@ class LostAndFound(tfds.core.GeneratorBasedBuilder):
 
   def _info(self):
     possible_features = {
-        'image_left': tfds.features.Image(shape=(1024, 2048, 3), encoding_format='png'),
-        'image_right': tfds.features.Image(shape=(1024, 2048, 3), encoding_format='png'),
+        'image_left': tfds.features.Image(shape=(1024, 2048, 3),
+                                          encoding_format='png'),
+        'image_right': tfds.features.Image(shape=(1024, 2048, 3),
+                                           encoding_format='png'),
         'segmentation_label': tfds.features.Image(shape=(1024, 2048, 1),
                                                   encoding_format='png'),
+        'instance_id': tfds.features.Image(shape=(1024, 2048, 1),
+                                           encoding_format='png'),
         'disparity_map': tfds.features.Image(shape=(1024, 2048, 1),
                                              encoding_format='png')}
     return tfds.core.DatasetInfo(
@@ -129,7 +132,8 @@ def _info(self):
         # tfds.features.FeatureConnectors
         features=tfds.features.FeaturesDict({
             'image_id': tfds.features.Text(),
-            **{feat: possible_features[feat] for feat in self.builder_config.features}}),
+            **{feat: possible_features[feat]
+               for feat in self.builder_config.features}}),
         # Homepage of the dataset for documentation
         urls=['http://www.6d-vision.com/lostandfounddataset'],
         citation=_CITATION,
@@ -137,14 +141,14 @@ def _info(self):
 
   def _split_generators(self, dl_manager):
     """Returns SplitGenerators."""
-        base_url = 'http://www.dhbw-stuttgart.de/~sgehrig/lostAndFoundDataset/{}.zip'
+    base_url = 'http://www.dhbw-stuttgart.de/~sgehrig/lostAndFoundDataset/{}.zip'
     download_urls = {
         'image_left': base_url.format(self.builder_config.left_image_string)}
     if 'image_right' in self.builder_config.features:
       download_urls['image_right'] = base_url.format(
           self.builder_config.right_image_string)
     if 'segmentation_label' in self.builder_config.features \
-        or 'instance_id' in self.builder_config.features:
+            or 'instance_id' in self.builder_config.features:
       download_urls['gt'] = base_url.format('gtCoarse')
     if 'disparity_map' in self.builder_config.features:
       download_urls['disparity_map'] = base_url.format('disparity')
@@ -190,8 +194,6 @@ def _generate_examples(self, **paths):
         'instance_id': 'gtCoarse_instanceIds',
         'disparity_map': 'disparity'}
 
-    print(paths['image_left'])
-
     for scene_id in tf.io.gfile.listdir(paths['image_left']):
       paths_city_root = {feat: path.join(feat_dir, scene_id)
                          for feat, feat_dir in paths.items()}
@@ -200,11 +202,6 @@ def _generate_examples(self, **paths):
       for left_img in tf.io.gfile.listdir(left_city_root):
         image_id = _get_id_from_left_image(left_img)
 
-        if image_id in self.builder_config.ignored_ids:
-          continue
-
-        print(listdir(paths_city_root['segmentation_label']))
-
         features = {
             'image_id': image_id,
             **{feat: path.join(paths_city_root[feat],
diff --git a/tensorflow_datasets/image/lost_and_found_test.py b/tensorflow_datasets/image/lost_and_found_test.py
index 3bb27f8b3bc..9600c30f569 100644
--- a/tensorflow_datasets/image/lost_and_found_test.py
+++ b/tensorflow_datasets/image/lost_and_found_test.py
@@ -4,7 +4,6 @@
 from __future__ import division
 from __future__ import print_function
 
-from os import path, listdir, remove
 import tensorflow as tf
 
 from tensorflow_datasets import testing
@@ -19,11 +18,7 @@ class LostAndFoundTest(testing.DatasetBuilderTestCase):
       "train": 4,  # Number of fake train example
       "test": 2,  # Number of fake test example
   }
-
-  # If you are calling `download/download_and_extract` with a dict, like:
-  #   dl_manager.download({'some_key': 'http://a.org/out.txt', ...})
-  # then the tests needs to provide the fake output paths relative to the
-  # fake data directory
+  # files as generated by fake data functions below
   DL_EXTRACT_RESULT = {
       'image_left': 'leftImg8bit.zip',
       'image_right': 'rightImg8bit.zip',
@@ -35,26 +30,32 @@ class LostAndFoundTest(testing.DatasetBuilderTestCase):
   tf.compat.v1.enable_eager_execution()
 
   # create fake files
-  testing.test_utils.remake_dir(
-      'tensorflow_datasets/testing/test_data/fake_examples/lost_and_found')
-  base_path = 'tensorflow_datasets/testing/test_data/fake_examples/lost_and_found/{}.zip'
+  example_dir = ('tensorflow_datasets/testing/test_data/fake_examples/'
+                 'lost_and_found')
+  testing.test_utils.remake_dir(example_dir)
+  base_path = example_dir + '/{}.zip'
   # generate image ids matching between zipfiles
-  train_ids = [*generate_ids('01_Turmstr_17'), *generate_ids('02_Goethe_Str_6')]
+  train_ids = [*generate_ids('01_Turmstr_17'),
+               *generate_ids('02_Goethe_Str_6')]
   test_ids = list(generate_ids('03_Schlossallee_1'))
   splits = {'train': train_ids, 'test': test_ids}
   with tf.Graph().as_default():
-    create_zipfile(
-        base_path.format('leftImg8bit'), splits_with_ids=splits, suffixes=['leftImg8bit'])
-    create_zipfile(
-        base_path.format('gtCoarse'), splits_with_ids=splits,
-        suffixes=['gtCoarse_instanceIds', 'gtCoarse_labelIds', 'gtCoarse_color'])
-    create_zipfile(
-        base_path.format('rightImg8bit'), splits_with_ids=splits, suffixes=['rightImg8bit'])
-    create_zipfile(
-        base_path.format('disparity'), splits_with_ids=splits, suffixes=['disparity'])
+    create_zipfile(base_path.format('leftImg8bit'),
+                   splits_with_ids=splits,
+                   suffixes=['leftImg8bit'])
+    create_zipfile(base_path.format('gtCoarse'),
+                   splits_with_ids=splits,
+                   suffixes=['gtCoarse_instanceIds',
+                             'gtCoarse_labelIds',
+                             'gtCoarse_color'])
+    create_zipfile(base_path.format('rightImg8bit'),
+                   splits_with_ids=splits,
+                   suffixes=['rightImg8bit'])
+    create_zipfile(base_path.format('disparity'),
+                   splits_with_ids=splits,
+                   suffixes=['disparity'])
 
   testing.test_main()
 
   # remove fake files
-  testing.test_utils.remake_dir(
-      'tensorflow_datasets/testing/test_data/fake_examples/lost_and_found')
+  testing.test_utils.remake_dir(example_dir)
diff --git a/tensorflow_datasets/testing/cityscapes.py b/tensorflow_datasets/testing/cityscapes.py
index f428b42cd5a..ce128c1303e 100644
--- a/tensorflow_datasets/testing/cityscapes.py
+++ b/tensorflow_datasets/testing/cityscapes.py
@@ -1,12 +1,12 @@
-"""Helper functions to generate fake Cityscapes-like zip archives for testing."""
-
-from tensorflow_datasets.testing.fake_data_utils import get_random_png
+"""Helper functions to generate fake Cityscapes-like data for testing."""
 
 import re
 from os import path
 from zipfile import ZipFile
 from random import randint
 
+from tensorflow_datasets.testing.fake_data_utils import get_random_png
+
 
 CITY_IN_ID_RE = re.compile(r'(.+)_[0-9]+_[0-9]+')
 
@@ -21,21 +21,22 @@ def generate_ids(city, num=2):
     Generator for id strings.
   """
   for _ in range(num):
-    yield '{}_{:06d}_{:06d}'.format(city, randint(0, 999999), randint(0, 999999))
+    yield '{}_{:06d}_{:06d}'.format(city, randint(0, 999999),
+                                    randint(0, 999999))
 
 
-def create_zipfile(zip_filepath, splits_with_ids, suffixes=['leftImg8bit'],
-                   maindir=None):
+def create_zipfile(zip_filepath, splits_with_ids, suffixes, maindir=None):
   """
   Generates a zipfile with a cityscapes-like file structure and random pngs.
 
   Args:
     zip_filepath (str): filepath to the zip archive that will be created
-    splits_with_ids (Dict[str, List[str]]): data-splits like 'train' or 'val' that map to
-        a list of image ids
-    suffixes (List[str]): suffix per modality that should be created e.g. 'leftImg8bit'
-    maindir (str): name of the root directory of the zipfile, defaults to the name of the
-        zipfile
+    splits_with_ids (Dict[str, List[str]]): data-splits like 'train' or 'val'
+        that map to a list of image ids
+    suffixes (List[str]): suffix per modality that should be created e.g.
+        'leftImg8bit'
+    maindir (str): name of the root directory of the zipfile, defaults to the
+        name of the zipfile
   """
   with ZipFile(zip_filepath, 'w') as z:
     for split, ids in splits_with_ids.items():
@@ -49,4 +50,5 @@ def create_zipfile(zip_filepath, splits_with_ids, suffixes=['leftImg8bit'],
             img = get_random_png(height=1024, width=2048, channels=3)
           else:
             img = get_random_png(height=1024, width=2048, channels=1)
-          z.write(img, path.join(split, city, '{}_{}.png'.format(img_id, suffix)))
+          z.write(img,
+                  path.join(split, city, '{}_{}.png'.format(img_id, suffix)))

From 347e3f1a1d16569ab86ea1a35e2345f0d095c139 Mon Sep 17 00:00:00 2001
From: Hermann <blumh@ethz.ch>
Date: Tue, 27 Aug 2019 14:04:32 +0200
Subject: [PATCH 05/11] remove double import

---
 tensorflow_datasets/image/__init__.py | 1 -
 1 file changed, 1 deletion(-)

diff --git a/tensorflow_datasets/image/__init__.py b/tensorflow_datasets/image/__init__.py
index ed9c567f9f2..f3344f86841 100644
--- a/tensorflow_datasets/image/__init__.py
+++ b/tensorflow_datasets/image/__init__.py
@@ -83,4 +83,3 @@
 from tensorflow_datasets.image.uc_merced import UcMerced
 from tensorflow_datasets.image.visual_domain_decathlon import VisualDomainDecathlon
 from tensorflow_datasets.image.voc import Voc2007
-from tensorflow_datasets.image.lost_and_found import LostAndFound  # TODO(lost_and_found) Sort alphabetically

From 46fb524e075d556d25cb7f4cfa31c295f9d8aabe Mon Sep 17 00:00:00 2001
From: Hermann <blumh@ethz.ch>
Date: Wed, 28 Aug 2019 10:45:32 +0200
Subject: [PATCH 06/11] remove double blank lines

---
 tensorflow_datasets/image/lost_and_found.py | 2 --
 1 file changed, 2 deletions(-)

diff --git a/tensorflow_datasets/image/lost_and_found.py b/tensorflow_datasets/image/lost_and_found.py
index 4e746bb65b2..58e0f9be2d2 100644
--- a/tensorflow_datasets/image/lost_and_found.py
+++ b/tensorflow_datasets/image/lost_and_found.py
@@ -212,10 +212,8 @@ def _generate_examples(self, **paths):
 
 # Helper functions
 
-
 LEFT_IMAGE_FILE_RE = re.compile(r'(.+)_leftImg(?:8|16)bit\.png')
 
-
 def _get_id_from_left_image(left_image):
   '''Returns the id of an image file. Used to associate an image file
   with its corresponding label.

From 02f23e818b8591985e7610c82e3afd6c56449a24 Mon Sep 17 00:00:00 2001
From: Hermann <blumh@ethz.ch>
Date: Fri, 6 Sep 2019 11:42:11 +0200
Subject: [PATCH 07/11] reduce filesize of fake pngs

---
 tensorflow_datasets/testing/fake_data_utils.py | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/tensorflow_datasets/testing/fake_data_utils.py b/tensorflow_datasets/testing/fake_data_utils.py
index c8506df2d33..970e1ec864c 100644
--- a/tensorflow_datasets/testing/fake_data_utils.py
+++ b/tensorflow_datasets/testing/fake_data_utils.py
@@ -54,7 +54,10 @@ def get_random_jpeg(height=None, width=None, channels=CHANNELS_NB):
 
 def get_random_png(height=None, width=None, channels=CHANNELS_NB):
   """Returns path to PNG picture."""
-  image = get_random_picture(height, width, channels)
+  # Big randomly generated pngs take large amounts of diskspace.
+  # Instead, we resize a 4x4 random image to the png size.
+  image = get_random_picture(4, 4, channels)
+  image = tf.image.resize_nearest_neighbor(tf.expand_dims(image, 0), (height, width))[0]
   png = tf.image.encode_png(image)
   with utils.nogpu_session() as sess:
     res = sess.run(png)

From ce9de6a317e3ee3d95d3a388b70a9172e41e03d0 Mon Sep 17 00:00:00 2001
From: Hermann <blumh@ethz.ch>
Date: Fri, 6 Sep 2019 11:43:01 +0200
Subject: [PATCH 08/11] add fakedata to repositoy instead of creating at test
 time

---
 .../image/lost_and_found_test.py              |  36 +-----------------
 tensorflow_datasets/testing/lost_and_found.py |  30 +++++++++++++++
 .../lost_and_found/disparity.zip              | Bin 0 -> 27666 bytes
 .../fake_examples/lost_and_found/gtCoarse.zip | Bin 0 -> 83202 bytes
 .../lost_and_found/leftImg8bit.zip            | Bin 0 -> 53484 bytes
 .../lost_and_found/rightImg8bit.zip           | Bin 0 -> 53509 bytes
 6 files changed, 31 insertions(+), 35 deletions(-)
 create mode 100644 tensorflow_datasets/testing/lost_and_found.py
 create mode 100644 tensorflow_datasets/testing/test_data/fake_examples/lost_and_found/disparity.zip
 create mode 100644 tensorflow_datasets/testing/test_data/fake_examples/lost_and_found/gtCoarse.zip
 create mode 100644 tensorflow_datasets/testing/test_data/fake_examples/lost_and_found/leftImg8bit.zip
 create mode 100644 tensorflow_datasets/testing/test_data/fake_examples/lost_and_found/rightImg8bit.zip

diff --git a/tensorflow_datasets/image/lost_and_found_test.py b/tensorflow_datasets/image/lost_and_found_test.py
index 9600c30f569..3e233126cf3 100644
--- a/tensorflow_datasets/image/lost_and_found_test.py
+++ b/tensorflow_datasets/image/lost_and_found_test.py
@@ -4,11 +4,8 @@
 from __future__ import division
 from __future__ import print_function
 
-import tensorflow as tf
-
 from tensorflow_datasets import testing
 from tensorflow_datasets.image import lost_and_found
-from tensorflow_datasets.testing.cityscapes import generate_ids, create_zipfile
 
 
 class LostAndFoundTest(testing.DatasetBuilderTestCase):
@@ -18,7 +15,7 @@ class LostAndFoundTest(testing.DatasetBuilderTestCase):
       "train": 4,  # Number of fake train example
       "test": 2,  # Number of fake test example
   }
-  # files as generated by fake data functions below
+  # files as generated by fake data functions in testing/lost_and_found.py
   DL_EXTRACT_RESULT = {
       'image_left': 'leftImg8bit.zip',
       'image_right': 'rightImg8bit.zip',
@@ -27,35 +24,4 @@ class LostAndFoundTest(testing.DatasetBuilderTestCase):
 
 
 if __name__ == "__main__":
-  tf.compat.v1.enable_eager_execution()
-
-  # create fake files
-  example_dir = ('tensorflow_datasets/testing/test_data/fake_examples/'
-                 'lost_and_found')
-  testing.test_utils.remake_dir(example_dir)
-  base_path = example_dir + '/{}.zip'
-  # generate image ids matching between zipfiles
-  train_ids = [*generate_ids('01_Turmstr_17'),
-               *generate_ids('02_Goethe_Str_6')]
-  test_ids = list(generate_ids('03_Schlossallee_1'))
-  splits = {'train': train_ids, 'test': test_ids}
-  with tf.Graph().as_default():
-    create_zipfile(base_path.format('leftImg8bit'),
-                   splits_with_ids=splits,
-                   suffixes=['leftImg8bit'])
-    create_zipfile(base_path.format('gtCoarse'),
-                   splits_with_ids=splits,
-                   suffixes=['gtCoarse_instanceIds',
-                             'gtCoarse_labelIds',
-                             'gtCoarse_color'])
-    create_zipfile(base_path.format('rightImg8bit'),
-                   splits_with_ids=splits,
-                   suffixes=['rightImg8bit'])
-    create_zipfile(base_path.format('disparity'),
-                   splits_with_ids=splits,
-                   suffixes=['disparity'])
-
   testing.test_main()
-
-  # remove fake files
-  testing.test_utils.remake_dir(example_dir)
diff --git a/tensorflow_datasets/testing/lost_and_found.py b/tensorflow_datasets/testing/lost_and_found.py
new file mode 100644
index 00000000000..edf58752da1
--- /dev/null
+++ b/tensorflow_datasets/testing/lost_and_found.py
@@ -0,0 +1,30 @@
+"""Script to generate fake 'Lost and Found' data."""
+import tensorflow as tf
+
+from tensorflow_datasets.testing.cityscapes import generate_ids, create_zipfile
+
+
+if __name__ == '__main__':
+  example_dir = ('tensorflow_datasets/testing/test_data/fake_examples/'
+                 'lost_and_found')
+  base_path = example_dir + '/{}.zip'
+  # generate image ids matching between zipfiles
+  train_ids = [*generate_ids('01_Turmstr_17'),
+               *generate_ids('02_Goethe_Str_6')]
+  test_ids = list(generate_ids('03_Schlossallee_1'))
+  splits = {'train': train_ids, 'test': test_ids}
+  with tf.Graph().as_default():
+    create_zipfile(base_path.format('leftImg8bit'),
+                   splits_with_ids=splits,
+                   suffixes=['leftImg8bit'])
+    create_zipfile(base_path.format('gtCoarse'),
+                   splits_with_ids=splits,
+                   suffixes=['gtCoarse_instanceIds',
+                             'gtCoarse_labelIds',
+                             'gtCoarse_color'])
+    create_zipfile(base_path.format('rightImg8bit'),
+                   splits_with_ids=splits,
+                   suffixes=['rightImg8bit'])
+    create_zipfile(base_path.format('disparity'),
+                   splits_with_ids=splits,
+                   suffixes=['disparity'])
diff --git a/tensorflow_datasets/testing/test_data/fake_examples/lost_and_found/disparity.zip b/tensorflow_datasets/testing/test_data/fake_examples/lost_and_found/disparity.zip
new file mode 100644
index 0000000000000000000000000000000000000000..131b093e9754018fd296c342329ee4d8b7076027
GIT binary patch
literal 27666
zcmeHQZA?>V6n;yAG8iiwhlYeNKL#Wl^t*)&uv?XnDwMB@n{Z)URUCtkmTfY@BAbyc
ztHdxv78j^L*i0r&Tx2f5aBL%SI?y4wFd1qPrXfT|hf_0mUZjxTUdT%Jk5h6Bx4qnx
zH1|FCd7kr}Q=OLYzKYOC=k{k(k6N<M2CX23ZVB+(>jg$beo298Z<MJdzn~~eBxW*8
zOWrh^N*J-yagUL!<SLPjktvlTWi&JQpOJ>5UDdjj1Yhp}Z#biGQlciE5HI-Cedd|y
z3UAdB;y03{QD?q;Z2an}G&gupo#%DY_d-OcVytWO*rf}R;g$ZPwO{$fe|^n6uI=xe
z)0HiQWh8^S!`sq+QXndQSxCq~^@Ny1c!a3o1P9#VBu_w~m2mF$goGp#vatpaz|R4%
zhvl{|OchP5MYfKh&b}VV8O@M0_F_W7#E3TtyisxhhM4$%pSt@w!o1rwV{}8t2tpWv
za0+21qGt{kVHvQwHO39_>&mu6#%KxZtQ8>nKoW*r6FIlD_RtXDCiJTx8hlF!@&`%U
zl%i>RA@6|BL4rB5=fRC&kYDKLsstlb$faTlqfkky663N2bIUHlSkqqQ5smu0eU)81
zJ%dvNWk-%0GHWEqFMDQv-|G=JROAu%`)_|tf6(#((9vNVY5bY$Finv89;Kqpf^rk$
zWyH%Yyc|;jtm>O4fymYltZEcl6>@J5?u}wS7A<J;#?lO|s%f^kV>^Yc3b{82_r?+*
zmO7!L+whCN-F$*ViV4UV*ZTQ&lSo34gdhn)62c-O6kbjVbfLmEZUJWX!WzuL&5sL0
zK!2f|77m+JQtUPu36oH)H|^Fl8Q_FbM2Q^t87Y`x#3J}5Q^=&Q>oB8s9Y(yBPl)Ey
znG+t*4x6q1*?UOBlUGASb3QTskg{!up=*EYxsT%CX$uU0&rYoFKo<PWM#%6Z^uC}6
z2)$5f`C>a=x~G9!RY1Y&e)i)MHx{f=h(L2Bnnbw(L&KkxS~Ww3X9#E&m2XFQxLR<v
z;A-JoEz8~ZF!u0tZfD<5vo&FAia{X;R|>8aB%)kIbV_jgil4H-aYOaOLX1!Ut`55y
zW+_69SSkSthLJ19Dw)D{A%<;;iL3gsA$aGZ&DYbDZ>xK?;UUk3M6As*+!%k!FnKZX
z+tk`}s>49Js;@K_6lv69IP^F4g*p4emOF4*j)1zmr;L};YG+o>$gD6iU}C_EK6gcb
zY@`%)m<h-hkx=3;hmL2|VbBnQZaH+zag90{Q=No<=T@L0)SnMDr15!ddS7l$?+B+5
zPW>l16-%z16Iz9>wSx=mFmJEi7hs<Ylj3NI(Hl)sA}N#c+U~+)qcOj*P_Jji&aW_H
zNwi$1Vw7@;RH}4cikTY^GuxXKSG83zrCy;CjeS0K(E5I+SsR^r+1{FT*QWWEj(VsE
zu=FYI&CJE|FepT$5RF1Kmk@1w2;*T&>A(l7+EU1b10V28;kr@~Qz53xje|zU7&>!~
z4tyw%HAeupKyraYFR+M%Lob|e5Z%*I<*~OVmAxX&I}eV2Z2d(K;c_f_a%q+LIm{FD
zUVio_;kZSfzGV?5W@v+d0JJ61ZHY>ZL?HoJ45Lu0<kBTfiMej?PO_%)VM<Jzr8;@t
zp*O;w*1x)O*!rVaa^mVQ?oDKUIoT6*Sb4q!)R!3AY(VK%c`PCK_TyE6LyECK7yENr
z{kgp(JECUz)J!KgHO64k3V9{oEbwN*b+ZVhV>dgXCus&;Kqn*eDrDbI>>INOW)H|7
zy?4(-Us5>~cZ{&rBecN5ASlA&M6m@7r0ryV-&3P0mcPz8sxUmaRs4m&DTy9SALdH>
z>@+ZDY2TZq(>exa&jzNU=)+uD1#*xpK^K`6%nr&PhVwWi4Jug25!7)Mjw_>a*n!#2
zU-N-$WUdEZq!o-EnC*6&54>ixL*W0x`{WdxrSHwhKE7>9u@_PMbXpHjdVWv%=>x13
IUO@N$1^cWOS^xk5

literal 0
HcmV?d00001

diff --git a/tensorflow_datasets/testing/test_data/fake_examples/lost_and_found/gtCoarse.zip b/tensorflow_datasets/testing/test_data/fake_examples/lost_and_found/gtCoarse.zip
new file mode 100644
index 0000000000000000000000000000000000000000..b6cc82aaf290f9403f57df8c82451d90a77bde6a
GIT binary patch
literal 83202
zcmeHQdsI~Q7CtZ_zIZ5>FA!9&nVHPMz>Fv;I$R?Gd6;h*NJpd)5QbDV8Sv`0GNZL{
z;a!H1%X^Kkg48n108uNJ*i}plyOL0tOllWJSw1qo`-gL&8Rig=`J=^rv_^zQ{Eojl
zzkT-p_V;}|bgsy17{@+J7f+a15_Lf9G=$^WD~!Guqn@1*os_I{S0_cs#=DCp%K0ft
zYm(JTN{LMWE2Wp*OD>iwr81dVHdV?0kILBiWOa1>3RU3B<SB{qF>i#1%&{FfY9Re?
z+rSqT;T+eW{%Y0wCl?xbbvMU3Gz2RA=dXY7my1QgmK>)Da}A2VF*)|V^zfzMe{p!s
znfjrhKVS0WaoLGoGuCZ#n)vG?`>1PWW6G;jd^PzOQaSGCV=MZN1L!vorC;X8ajq|L
z+{{dl)3{o2oIm}_4EzB00r!Ksor}ENjj0w9b=NI|d}~CzQ<e5nl_CTL1V%g{@SyCN
zFjU=-Db?iHv~d3_`<9}`QbY@+QAj6|RvWJ#s&1($TAWGcB8n+uY^MC88AS(*FBGAu
zwT+Y*cWi8K+kN)Z@r4Zh-~t0e6a}-QGD7vBm{-cie&b9q410ypPw|LK&!L#O=$BP-
z-N0e4@HmVlc(w%>aQ^E}L)><4dE0J;Yt*5Qj!tUTm)7H_JlM8zXNK%@Np_8+s)<Ug
z(nB`lmQsRaHu!SfdOsj8K%RiW0f{s~WVP!L4i~q4OktvlQl`6<GEE0o3(OkOJX#b+
zS~bqaEtah)%&~)%TzopFJj(`m3hopmcM9MM<M4zTf&6(Zd^LxzQ&y2i)yV4<Fn+Q?
z=NHf|pj*Voj<HeeG9kcWO6Kl{VqqZW^w!#<F$BV}*IbhTF)I?{5|X;ghWUv{Vf+b&
z8KyY@=<3q)nO|hq4|sle)K*R92UdwQ%L8K8o(mK2$*wJFSo{8sSBVI7PKQDtE1*>A
z+%qV_QGy$k;CLk9k&sh&-8$%dYF7F1gcuZ7U<smeiN>Xo#%12^1Zr_-Q({q@c8L{O
zet>KN*#d|I5N8C4bDqI63mtLp8bLD8x)=#5*g7LyhdRAuosMFqPZukXEDlJAoJsuW
z^tXgzn9!^b#XJlX);$<T>h0wz@lblpJ()s@C(`spn$!ly@Y19mQL1r+<hDolESCSR
zNagb4sM*(_j=WX3r@k_G+GeMTpD{MfN7Uy2mGY$nEgV<>7{?pO8>}d>q6{I~iZdim
zI?fn$1vcUoh?C4sorYW(xp0@o%2oH*5Ehe5Sj-sK<t{{*yUSZ=MpOud6oeFZF0>Dq
zxkp_~rnYb)MT`Q97#j?;54a+r?}NS%`aT5GJ`mlz=<xE5wB=ib!5G`Ei)vhm55r#F
z6ekIQG2Im>UFB&pl3+4%P@MlHf51tFW6*rd$cq<?PKMhq_j&mPyTlW=kA1CWk7{pS
zpZOhYSuyj6--r{Kvc#U_9_7LV1XUKAm}p`eG%?jBg``emo>D3ul@98pjs_Vhq+lEa
z;~35=C#`1C?gz|;?h}d_tkq4q1!Af}Oa(pzKSC><3;)J!(lYL4Z60b3r!0r76sR42
zn%aRg-wUhNDi?{1-OYo8ZAJ-0F{7I;#XJ<#O?i?46w@tvQacV4MkX-|f4?o;UAGP$
z^e&e%ZoG%MJjkl*b5Z~8g9_GvHn3n*a>XS_6=M)H44U=1TS~@o+`}2*Lx2x~jszrX
z2Kk}(=pn*k$nT+*KBq+A3RNYlN;LHan)<0zS+roDQX2@hJx*B$P+Q<oAU1&5V2s${
zd|W4eZbfc40=30+Mc}!D9tWNx_T_sV?qrUxoh*#Q<V+Zx&+C(3GNMoNP|itEsaLC%
z5i~u?+g+^xwbGLYCrQNgN$M^2G>t|{qC;dHZTsiVnz<^Gdu0ok`exAJ{X4Rb*1uW&
z@A^vFx{~9euMr5{+PZ2af=uCc3$FaJBS;Y_wNPr|k!bKp3~8k16;m86CuKKFoG?Ji
zV29A|=ta8&jH158D5|<u99dpnLrKL9UAr4joPFv2#c9R6X}a22tEPznO!j<@u$cTh
zptl&DV`Oj+ScG5^vJ~6b;-<5GMv>F>5@A5b_CT&156Fa@43O!rPN~f?Mu*0n+<(Hl
z-{w92WLfin7N5wQ<g?Fp+>hfHmMphkSoZ0KmbX7*PB9fasZwbn`EeG)y%#q<>Ppm=
zhHAx}cL<G{LbVlBG-50YkQih~cu)vEDDrMI#~4y4<-bF<)g!`hn~k-_;Z6s4I<T?L
zo{g=3u!J0A)={)zrO0D2Z_y4$I~eU?ffPH{^6o&pv=({89bq(P+^Pi%UZyn9WN1w9
zq)IER(uX@KtZeuFIM!ozW~QYme?(;)Gv=XR*lP@qX0}<ylrrtD5pkNwz+#YM3#HhI
z4+8PQbX}r9tynn6?^`l~h@^OtK=A?ttx-jxiZc5uN)V58t|8?1EMvtm<n{;#T7ydh
zE{Pdr2|Yqr^BPi0HTjnahhfeMn}OUyCvTvW|I=p0J-XjfM)jtQy2Fj#J!E8Y?n?#2
zSj?%TVJWWU+{<1)kSa+$JxHXa^pZ*BQtw_#l^WYje05T#*}F0<MT_3BeNnn{-**35
z%c{QQCaiAAc<ypR#&dPGx9%#wYoa*FoFJHe7TjY5GfbR3RG?;FfoktXQV*+y{2-X~
zHy{}t2s1wrNE178T6E4OKM37ufT(LUvN4OOx8UG|gAZs5(3BxGm6v^&09-$UVvez4
zjY}bf#<Qi9XA6okvwo6RiSDhl-#2n%#M8o%jN90?qj`Evc=zcs0+391#Y$~vzW%|j
zX1*_$4X@uOnx2`MKVV0_wR>>p`Ms0!rd&L=C+u)Vwr@`E#Er~5hB0KAYUv!jV~{ce
zz3mNp+lTSIotJ37bseu(qTxy?usB#VLW6>tlQeU(LxI(BjJ#vm^fgAkWrbEV?iAc9
zM(z~k8^|}*51I2fVYbzLcNp`nJH}L;6r7YkTwhpql@yiA`8zn+MFjsnUl@+@Sg<32
zhhuu6RuX_?dZku6xykk`*9>o4Ra`XZJ0dT5q~kxTC$<e%I&gcC>Npiw%WKG(lgSWL
zd^zr38r<(uVj*!x;%tyOKe6UyW)O@T(?oFt2@?{gE<ZsHt94|SeLA!B@uLCP$Cy%y
z4Y)o>5rc*{c7SYf6ZGku0B0UkP`Xr2gfMDMJf#$`GlqT<`aB?zfk1}nX*_!7J@agk
zL%}ZPl`n*`nDgs$&W<9b5_>JuqsFL`)$U?XWyFfraS6%E(Q$Dql~U62J4%ViR4=Jq
zDf9C1^py2Vu=E2jSSlidjDx~I?XBIfEgNVp-{*ctGIiJDi`%DMtpCsa!>0r1t=qJB
z;xtAOq`axs)*O?rIm7|P0poGN+=~b8dg@@MFf!e%G>P(V-Bl5Y_sE+NFA*<0#>=W(
zi%8eVyg8YZZy7awnk~R2g8%^n1n4X}>MXRsyk4rwXQN#ildjc02GPP5(E`x|(ZWcy
zu>Y@UNZeCLDj3kh1qOsD3T8!Rgw7RN4P)QnN*7O<SFvo_1%9rQzI#kB@tE$amD)!u
zF||@am|yfkORGGKY1UJps<Ks0{>#sbjBUs2ZSOi+CzNccsBly<^+689WD4OOgY%E`
z-|_t8x%lK<XxA^L8jCTC89}CzGN<VrP7F>AP7FMBI@%gP@s1%Uh?b8jVz7zVAjsgP
z;G}dsDR_c#8Ig}@Q~>jhX(17G^&0B;*mqUB064rO9F8^-+C+W$T-ZCuI8+2?4Nhh7
zi*Z0EEpN&Dyq#}2{Zx#4c0zPgvdUfEOF*VudZl(R-(YH9E6#rsn_;QQ^iZ#uV>!v;
zMEH=F2rs#I=b-p|P0t>^)wr|Jrt&h=DlMj-H#3mAK>xs-k3m7lm}A2+$H*DXK9_zn
z0-=dun7sty{zkdePiSJHi3LawV@ZuTu4x|@MV42e(Y3tGU>*al77A?tE{S){F#lF>
zK1k*<OxeinrkIZ<z+1CVIL)1pjy|wJxAzgZD#gxyy61-)VLYa3|1O`ABvxXto(Yyb
zygg{dlG0ly_wtnXO0ZPV3zj5dqEJrZ?>B1osZi^=t{!XDd6uI2>b39cW-HxKd;dlZ
z1whupB1}5JriJ6`@0;QCzzs4b<cr7`JIxpKZnFxDk%|*^LT}(P=)&~jE)0sNUiheK
zmyv(W0~(TYja6Bwu+ZW`iwEGn5O{yKiRJ`k>y(X*sgKkGsRdFC5TA_YU|I|NWA@HR
zX8rJkFdXyl!i2!y>6HZFm>%nuiUM2pO5-i2*_`el>r_4~aa+*j^g8R>lV3mI{N~29
zR)r-c7b*s+X0YOlX_bm8Zz3TAcXN}mO#oOY0M-G$LaJ#Vsb<yvHKm$-X6SpDGAHJ~
z2SZ<2lp_8i{uzsZK51-T$|4=}mMI+3@<z)WdL`(UjP*W4Ds+I%|G>RPXB3^$PCKLK
zP8`y@-2ckqX_qpaCSgEkz;EaJITJUAy?Uls5&&enrB}Mr#*A4*po${Q)ouUWg;v#E
zhR?b=;VsqH{T#}}hu7Y$i%4uZ=bUR($%?8{oq0^k5|V6Qrom8nVi<d3pzc83VXW>j
z6BOSkQa3N9)%Tl>IwA0en=xeS=wcd;e~q~`FNL}IwlZXx{Wt6&@c1<2$EVR#g1+Yy
z0$$@g6fOKHTBNeMB4nOWiAez{!cc?(m@@**Rn2fXY!|<AMd9cAV2p+3Fj2doZ%Nbv
ztrI~p>=nj-Q1_upXMbrolSzt>jdvGIl=D-P)+DQwloFZ#S4uCrms~7WN@X&!Y^tDY
zDRi#>@~@PQ{f0+_Lin}mvnKs5x}e@K|9h5>;o7IS))w*9pt*W$4;ycawc7wG!-Ye$
zJ{0o{>+ks<X}1=ZdV6_FJe1yYPqy{^ZYg8qlhx7jD^!6ilcyxc#~3beyLC|wKi*t_
z&nN!X6w7zE=cAh~#r)d(d;al4Q>-nx=Q$Gw=kt5ce5?5nP5QLhjthGz=On1qt5wPf
zy4Ahi#rj_>J?So&i0PBmTk6^MeLs-v#$N(MKKTz#`d7s6;@>@Y^^>EfzhqZ0p|iao
zw`zfc-+MwY+deeuOLmxQeZl=dbu=u6x2e+K|EoFOUR>fy0!gXTOD2(%P?F!dKf3ST
z#;zUB|NHg#e?y_EE}^si_gJtafL~vK|NEaX)%t?_Z#YMffA=-%SzrHt@m`vS{Xyt4
zbdPVp3b!jJyo%1=1w+5I_DeWIqF4T2GM#&xVe~cB2e+>uU}MpP^#!^ifAB^a(qh}M
rqI!f0ucEUz!mtt4ehI2cCcK2-oe1lGY>lDx--G@f*F23l58VF&Zu@*R

literal 0
HcmV?d00001

diff --git a/tensorflow_datasets/testing/test_data/fake_examples/lost_and_found/leftImg8bit.zip b/tensorflow_datasets/testing/test_data/fake_examples/lost_and_found/leftImg8bit.zip
new file mode 100644
index 0000000000000000000000000000000000000000..813b5942b1200ca4d27b557a3a5671a735cc13aa
GIT binary patch
literal 53484
zcmeI53s6+o8OJXVc|}U<`&Ays$Fx=1_pS{-M?(-{Np$dqxl8Q|MqUC&6bs8r6hoq|
zwAw^vN$bQ`(`cfd7=#Gb=%g7%;~+#75r?Q%BMLSuF5B~+Lsf{ziFRZ&{CDPZxy$Og
z`#aw`_dDPDpYKG4ySew0=v6Xnn9<iBcHrwENund1y_aB~pB|PNXZZDk^x*W=*ab<!
z3Z*=KacW{(daA6{x4)9L2CYG%ku`d~LO)u*@86C}Ns3z?6&dR5Gr))aufAa~PI_6A
zyx6bql9wwx-8PTvDM^7+*rW;5Gd5kR35pyf^_u*W&$P;+ZwBS;wr<HB-$OGbaM<6r
zCyccAp4!cI#n4#q5iZa8tU6Hn=0sDm_d46D!d=$$$+cVaODp?G4%?w`_h0mux*JZ%
zL>eVinXPgIT~-(Rv(ngWd5DYn#p*83UOmJ)@I`U91mXik0vuouL;&qV*a0|z0};UU
zLfAo9;V^sCm@ARV%#--Mse(L-nYL_Op;Tt6Id|rGPpMyBo3}85ycza#_ZH`8hlum|
z--)w?4+s$801k`*+J!I!Z~zA)fais<gRa71_stJ)g_OEjT=ZWwG`mY4r?N{ol3Q`<
z<IUt{GzuCRgo6VSz~EEZ0XTpI5$Gy|PsFCVy`#ECvwt*)*_R!~g|*C&*f(Pbg(n(Q
z8xxngN^U1Y{Sd8yla1gMZ%7EziT(r}z<~%jM@+W9JcVfiT;3%u0P4kg>ZL<nH2Y1=
z@&AkWC%>IRe+Zi#Iq8FmCTmnXImvT=mWcWwNuuM2%Skjkty-y)bp|!5N}fbc^7kqJ
zQooh4<CG^a_t|Kj7i#M3VI928=LO~a)syn)d4BZj@e=di&+Co4yamI9N>VcChBh{D
zjJtL{<5Tyts<cIR<ro&B@m#X+z2TC=WN$e_k(v=kKS5-yO`M&i`G}iAmktg@z&W~f
z)CKASQ+<#E@P44HOzk|hx<{UHBpOLdVsAP@p2VP1+f>r19=2`AoDX8U5G3<%0>W(J
zHGu;-FanSO2r~eOhjEx4V?S-9IiJJH`?j(`Mcu_Ki)qg1NOS<jEy~=`j?j)EA;DZa
z?*|?|hYU~v#R$Ov<Yv%iIm9;{x^}YJgGp0g@#jlvYY%=yQHzk)kj+e~Dz(jmpg$32
z0uJCn1kh~?I{*i8U<5jE*dA2;b~BTcY!-47D9434Iu(Z^$LzYm^=ad`ZKj)hu#{v^
zsNH&&$w=rhvQOP72l-!S5|td9Y)+3i%hOnL5?!#O{k5!Sc}bKCcGBoH>L-$u1Wg$#
z^;&ss>3+AwXyw4%*mV&po;!Y-SAP4{!7Xu$0lvfZMVUKS_v>Dq`}2a(n1ioZq(wKL
zZ&}ng(=%Y4bTjLf=y5zgQ{gyH^F0}J1IVwkToRrXlpDfA!2ukI02D034!{8%7y&F4
z7iItsZ5)nXZ?&JH_{<&qDJ~;9QNNU?eWLvi@ufbP_90N6H~zq?U4+C$Qw0ZbAOZ-9
z2|EA>C*~0UVOr`~N=&lPHi|EM$#lF*ZpE6~B(5g$6!(r%Iyevkj826efCD%%0vMeN
zGXMv0AOd&S{2WbfTBs$rV#bmx3Qz2;E!{-<&T0<r;%i@A1TS=>;J^r=`w(UT4&Xop
z?slWdCe00^Gup*r<4a6VvWd${F!~kk(y2H!ck6HTPkO7aB#&<H!D^B($Gqcro+(M_
zcsw<UQmtY-5?QNP8Z^2mQj@$p<!6+fWbL4XZ_GUK(N2}G_4IFIuD!56?hpT%YpK=@
zX#T}ikAl4Qy_dMni<tS9cWrLt{!2|U<yG>hpB?*ZgI21w<vS|KmrAjpGf{{pDXX=R
ze5xB4g*R1J!}96S%c(>^fo-lHqJW}uzyTZ>0Zdd1GXMv0AOd)12|IY29Euw1uU3+z
zWckTk3QB#pHdE$`)jLgRY3WafxTVhUzMaa`d}xLj<X7bug31yN1suSE2q3C0>;N3P
zCWn!>HOX7`EHg<`OAYxEr?Qv-i2@bba|R+SJ%YRNns~PWZ$b2q-~bLp0N#SI18{f<
zhutxWg?ww?ZyLBn^*@dhz7S?d!Bw~wa4X<e@OFd8spvky0UU^cQ#?-7BLB)12rhpn
zH3@c(=Z5JN_0q8}ntkdj0`+Uh7xkli{H~^?bk;Cq(B+ba4Xiflc$s~wU4{~KT6(ZT
zEl-;lpOBoE7MqY@Hp|NIz9TDDqqQ1?tk<g4YW<VRN+Ou7<n?1;2Ku|+(0H%wpW!tk
zC)eA{C#PKH;#%&Ww{4AOu-127x2Ox-er2_k^%#<UebEv3s_ah|uFC!(OKP-V_TLmi
zY3Z-NwUwl~H}`%!mga)SZEainBa?83y)aV?4n*K-&QL-;#BB~u4^0nE5Bos>IJAQ&
zW^CASd?3@QHebyko$A=6zI+MLV!MGH3xCm;p>YQXB7kwaumf-a2O{8<<6<b_xN+i{
zCE94Ga9H@od70dbolV7jDe<o3`4q3%Q=LWP66B{BRrBL;5wZ6juPsy-jFu5Vct@B4
zIDi8Yc$yLBq8kMp_g?g6>DfnTZzPeb{WCjhRufufnia7i{~5PQj7mh)1P5?n1h6Dd
zm;pF+Du>hl`N2$9QY7}gVYkvL1M`j|3IvO|QQWPIqr(GU3(zkboUw)8f3T>;+VXbg
zMJ6br<FQ30DxHcYE0J}2gI2A1B2h`ulmKba%Gf_Tx?h@q&h-AQC7$Ka=k=MDZkqXQ
z`7;^T$pv{HE6Pnj%`Jb!>nG8xJdAZia|22;3wtiNH_RD(yUFZraa2^Tka@O};(dCW
z_IciDqqW5*^D<6HYHURifRjKFKo|r#fCD3dAb>CfZ~zA)@HB%VPA#{KzjeDYVVRrc
z)|$1PPXHO`T;dK!>e3MCN-%FMx(jds2O@yTgs=l}00$y)HyVMI!rUP4wTt`>XB_WQ
zwvsz%`Lg2Iva@V75mdD|8Ow@cH&|B8+Y98W90TOai1u?sz(lZU58&_+4!f(XOKi2w
zkMOI}^DsrryhOfVuPvBwtcU%22^^xF!hFC19Ebp-y}}N_q0>0@?|o|m6P0Y?q7qEX
zih6KL4i|j?Ic)IUk4^s?M>ly-SmNT^%dHo`?-G`;#Md5n;A<u;p(9*;8~)RhL#KAx
zOHyMOBn2y!^7O^2iD~JnvQppvO4b^*28Bk}==BQy=#C1LsPOg$Kj&wOsAq-J(Y|2g
z;2&DBMyFLPRkF^Yrt<H*=}{?3aUBcZ6Kc1fWrfnwmb|3x8^1H@fd#AN&}4IZyjh;c
zzO*h_(f(Riv+t`^uoK$}Nd2h7J8n6|OJ+YTITz+XPT}9X0nz>6{6V>C?iC)gx<ldr
pmrOMs%kL;-x>x@1w?Ah2-z#d03h(AYWw1RAq|0n%`tQSZ^<Q6u-oF3<

literal 0
HcmV?d00001

diff --git a/tensorflow_datasets/testing/test_data/fake_examples/lost_and_found/rightImg8bit.zip b/tensorflow_datasets/testing/test_data/fake_examples/lost_and_found/rightImg8bit.zip
new file mode 100644
index 0000000000000000000000000000000000000000..2bbf45fb5e82127e3b6ad13eafe2860d0a206b82
GIT binary patch
literal 53509
zcmeI52~<?q8OI+C4yd@}5)lQW=0p!L!>lw0mqTJuK@9Y`g!YZbVX*=RP^b_c%`t5>
zE@`b;*=$p_CtCL<39==|mQsT$$L$bt0YelO4K8dWz2CixvNWDzo6ITyIp^_sa|Z66
z-}}CMzkBchelKLI!ofwNSMKc5(^I!q9=_@!Npu9U_u``?qY?vSB8~H+6MYin=SIi+
zsMK=U;`o?^#CTb4XniH?j5?!AD{Bn~mBCMb@{he2#YVmt5<IzAkD)!-|Lqm{TEOd)
z<ivh;keuw<sln_$K$6_0z<`Ni$r(4wJc5UXxV$?<8KDR|XFv33XLehbI(s-d+n0_C
zOH%xFp!=4M8#b6!?)@Cz@fhheDdM^F4n;wEev#esmwuCyblZBZN-0%Wl@={nYA-3W
zdavL9f~2|_Q&V%tK~gRZ?!yXXFaIoYmL^gS*vp~2IQJbX&J#9>Gd=(gL;&_6>;N3V
zfe4_T6m|d(U4w(yh-Ei|<0Ml-ap5_NB$`s1e)6)N#jfg&w9*06kc$t)h?Nckhf{xX
zc7IKrE$(*W7kmI57y;M={DKIeeGqm44&Xopx>~;(Ro`f>y`z+r;i)fWWl05=vLl&S
zyGzc-x9>W-nB0D8<zWN71K!aUy#v8Bw|lrh+#l{QNf<i|KX^KaL5(G4rLNMDMfFun
zX&GeM*Qvj!=pyHol_oDf7XJr>P;eju=wE~#fCD%Xfvz?PwV^P3Ro=NBl4SQM(gL7+
z67Fc592)u@jPML7YHGMbwG4bEIAHfA)5?$*O-TXMlpKmvZLef{5;{7eDbeb68nv&i
zH)=>)@=ThNJyYDIVeigu&hS*`nEZCRzP~f4d-eQz`(G>_U4HDN+TQc``m1Ig{dKqB
z4t96!7B{J`u(~$9@Qj>y*XN~To#dZgeX5A-N19(wr#MQM^Mj_d#h~MN#4Wm50r?0g
z0U3=j2yg%gA^>@rumf-a2SxzuF<}N1PF}uQa+|OFjPu$-0@cMe5j0V=PCt{&Ngk?2
z;5Kj@xDCXH7+vso08zDQ4#5E&h(NpW;9%YT1YItZrIt9Zr#NEop~e&nD`L+yPiONL
z7DsUFkQ+AraKnI@UHAt$fCCXgKP&729J)4#lb3IN*GRJ!N5kEMY2D+&=UhjU(QM3Q
z%OTRkO|@7z6=nhs;6MbPuG7M9F>WaUB^(AE{udmKe{ptZnv!hNl+gc&%u7(VDGpnc
z+=jXLe-QffdsMlO<s|6~#*|%SQW835uunay2zivE#8;jiXHJYV%b_ewiQY%m`dZeo
z1SM(}J8AV=&9g~Le5MSSh8=%P{@~JxafPehuR9IOG5>AsmQ88%O}@Q~|GCbj?$L8l
zq2lNXH<Mx3?0sHQclv+UCvTbm3&{;>k#D6W?o$>t9;mg>r}-IwYrP-sCcfaGN&%JS
z?p3h>R8Yn;qtMHy?v?w))hRSr;^u$@I4}aXsnbIu0&biblL;RMhyQC1Ti53ND)(wH
z$$8fKX0BBAS)Wb|L5r3o`%{21Tzq>>aDxL8z{IYw18@KbBG7I@9Sg(UFwkEJ|LC$D
z)->n)^Ymw@Zx+)=m|dG1DXh3v(QuzGlLa2gd%=MSbXhk+GGXoj=tf0501n{52%uXO
zX3(C4SMS|fI}B|5%bmJtz77&6bA8E$^eYG}Fgizf${h#Y2)Yq;BepcE$lj71%_Jq;
zxTK_gw}$On++l5ERyEh&)qZ&2PZwCeq&jshESXX{t+s-RO6cgMutcr#Ws(wEXHXlp
z`ezfCOdK<s2}>^3kM4Q+b#K$Qjr%?S-tGGQ=U*J_s!v!k`_nIy`V9K)c)zos6mIOf
zsQcYL7si!zFSqVVx?OYqo550Zw)K1|Wsv>YnqZ(6pR$zYe1q}*N2yAxV4b#Oowygc
z7mNTRJZ=WyATWRfI1mAwh%8TDHd_}jRY;0*!%ZGZR9=dwxti1En~<u+mIfuZVJ>)X
zB5%8F!gX;ka4#4EEXQ&)00&zdZ64fzw~nVOnOS^-&sP}3`PP?;9aVIBOWZPO>cN2s
zVE>b_18@KbB7kii!VbUz9Ed>67R)(c`@03vb9$@3dy{4<mh}FTCM`^nSNMZ)#tB?#
zM!|syp!pDX01n_l1X`L=>>B0<@q_!tiJ~ORRgxugm0<KM+|sr<<k@Xab6aw7!*V6n
zu46^Xidk<AxWv>XbWqZn)~QO&35h-`jU4)B)WW!ggt-eBn$5Dh{X4SS*H5Q4$_AaU
zMq_w3QHjTtQPPlAGk!NSCREmZWa|Iwv5~tc6t3NTXr1e3<E!bWfjh6nTzYPo=K;G_
z=Z+Qh$*$UU@!tBJ`ifhz3QJ8@P0?RTkJ~gSlRx&RqPm3UXwGhVxShCZIh2~*Qn8ND
zF<|{#U;qbjAOctn6m|d(;6McW9z5%ppIy!po$Wbwj{=LMd&YBV$vmqHSxV4r2rmH#
za3BKcHH00212_-?$a#bvwCAv<;z%_mKdY;KFplB~$NV@RR$PnLA*?`mD$E8Pz<~&$
zI~8^S4&Xop+IOdz)aRBlagDX2n8FHU?m3=Q^?Vx7PBP3(rs$$TA?ychr{KT{*w$Jl
z9x%g=621xF?7F`Bi|4;eWulS{E-LxKB`s(YpRP9jcEkr;J*Ju$OzTa4+Oec$*V@Bh
zRWdmV9i5ky`09OGq7qqeFzPhgXOonuy}6_$?vE~CY#cf^d7m_O#y^w+ANF(IYd0f0
z$YXVp%am;!w(eDoTD3ax&kN`MW_*3dft6*;KUs6*osxl4UiOg#Jd6s@JUWe@DOh{@
zCfAs3hNy%$))<#@3>c>)2RSzcj8BC>fCD%X0h`FKW>Y(g&(12#AW=z0F%K-78gn+&
z!}O{ft=LhFtuB}f;q4}-LPYlg4&cBDV7gYA0XY0S4s)!B9;8tIvwQvz_(VlT_2)EE
zkulzla@AWL#UCOZ0|z32EkME!zyTZ>0nB0wGXMv0AObD3cPB4@yCjmr3h(mMT%;Ot
zrHY3Yc^7DpEy4<6FX%?Vfe}FSDa-&Iz<~&~bR$1fir>YzmlrchNw&yS^5fUX7MR#v
z9Zn8@d!XlyhRkWBsA`W@C3f~M3Kx1y%AGxWdg`{y!&jN6gpMHouJ~bDLFd+qOXBB7
z$NH$$a@gYdn1sZ5S#4;2CF_hjqe?4l4F;9LubsRkWa{Ih3s}*I;#AuoNm^PL9p?ET
z7p>LnG-_X2Z`4o?o~*a`qS(l`MW-(qQ+ADgV>;-OKaXCP6(0t4T(qw|InJCIWtKzP
z7uWl!T3^c=_MO!#c4ChK(mb{Jwoe>-TzpeRr;7jn=MT00_HEPCJT5-?y|%^wcaqh#
ltwCEg)8iWaVP|J+@cr7Rkg45x!XBlhl<JtY{r(9m^IulH(I@}_

literal 0
HcmV?d00001


From d2cf70a58960085750d36517aab81d8d3c921aa4 Mon Sep 17 00:00:00 2001
From: Hermann <blumh@ethz.ch>
Date: Fri, 6 Sep 2019 11:52:34 +0200
Subject: [PATCH 09/11] clean up code, address review comments

---
 tensorflow_datasets/image/lost_and_found.py   | 39 ++++++++-----------
 .../image/lost_and_found_test.py              |  3 +-
 .../testing/fake_data_utils.py                |  3 +-
 3 files changed, 20 insertions(+), 25 deletions(-)

diff --git a/tensorflow_datasets/image/lost_and_found.py b/tensorflow_datasets/image/lost_and_found.py
index 58e0f9be2d2..432447a85c4 100644
--- a/tensorflow_datasets/image/lost_and_found.py
+++ b/tensorflow_datasets/image/lost_and_found.py
@@ -142,44 +142,37 @@ def _info(self):
   def _split_generators(self, dl_manager):
     """Returns SplitGenerators."""
     base_url = 'http://www.dhbw-stuttgart.de/~sgehrig/lostAndFoundDataset/{}.zip'
-    download_urls = {
-        'image_left': base_url.format(self.builder_config.left_image_string)}
-    if 'image_right' in self.builder_config.features:
-      download_urls['image_right'] = base_url.format(
-          self.builder_config.right_image_string)
-    if 'segmentation_label' in self.builder_config.features \
-            or 'instance_id' in self.builder_config.features:
-      download_urls['gt'] = base_url.format('gtCoarse')
-    if 'disparity_map' in self.builder_config.features:
-      download_urls['disparity_map'] = base_url.format('disparity')
-    # split into two steps to save space for testing data
-    dl_paths = dl_manager.download(download_urls)
-    dl_paths = dl_manager.extract(dl_paths)
-
-    # point segmentation label and instance IDs both to directory for ground-truth
-    if 'gt' in dl_paths:
-      dl_paths['segmentation_label'] = dl_paths['gt']
-      dl_paths['instance_id'] = dl_paths['gt']
-
-    # first directory in the zipfile, dependent on feature to load
-    sub_dirs = {
+
+    # For each feature, this is the name of the zipfile and root-directory in the archive
+    zip_file_names = {
         'image_left': self.builder_config.left_image_string,
         'image_right': self.builder_config.right_image_string,
         'segmentation_label': 'gtCoarse',
         'instance_id': 'gtCoarse',
         'disparity_map': 'disparity'}
 
+    download_urls = {feat: base_url.format(zip_file_names[feat])
+                     for feat in self.builder_config.features}
+    # Split donwload and extract in two functions such that mock-data can replace the
+    # result of the download function and is still used as input to extract. Therefore,
+    # fake_data can be compressed zip archives.
+    dl_paths = dl_manager.extract(dl_manager.download(download_urls))
+
     return [
         tfds.core.SplitGenerator(
             name=tfds.Split.TRAIN,
             # These kwargs will be passed to _generate_examples
-            gen_kwargs={feat: path.join(dl_paths[feat], sub_dirs[feat], 'train')
+            gen_kwargs={feat: path.join(dl_paths[feat],
+                                        zip_file_names[feat],
+                                        'train')
                         for feat in self.builder_config.features},
         ),
         tfds.core.SplitGenerator(
             name=tfds.Split.TEST,
             # These kwargs will be passed to _generate_examples
-            gen_kwargs={feat: path.join(dl_paths[feat], sub_dirs[feat], 'test')
+            gen_kwargs={feat: path.join(dl_paths[feat],
+                                        zip_file_names[feat],
+                                        'test')
                         for feat in self.builder_config.features},
         )
     ]
diff --git a/tensorflow_datasets/image/lost_and_found_test.py b/tensorflow_datasets/image/lost_and_found_test.py
index 3e233126cf3..0f804e14a27 100644
--- a/tensorflow_datasets/image/lost_and_found_test.py
+++ b/tensorflow_datasets/image/lost_and_found_test.py
@@ -20,7 +20,8 @@ class LostAndFoundTest(testing.DatasetBuilderTestCase):
       'image_left': 'leftImg8bit.zip',
       'image_right': 'rightImg8bit.zip',
       'disparity_map': 'disparity.zip',
-      'gt': 'gtCoarse.zip'}
+      'segmentation_label': 'gtCoarse.zip',
+      'instance_id': 'gtCoarse.zip'}
 
 
 if __name__ == "__main__":
diff --git a/tensorflow_datasets/testing/fake_data_utils.py b/tensorflow_datasets/testing/fake_data_utils.py
index 970e1ec864c..076a161f81e 100644
--- a/tensorflow_datasets/testing/fake_data_utils.py
+++ b/tensorflow_datasets/testing/fake_data_utils.py
@@ -57,7 +57,8 @@ def get_random_png(height=None, width=None, channels=CHANNELS_NB):
   # Big randomly generated pngs take large amounts of diskspace.
   # Instead, we resize a 4x4 random image to the png size.
   image = get_random_picture(4, 4, channels)
-  image = tf.image.resize_nearest_neighbor(tf.expand_dims(image, 0), (height, width))[0]
+  image = tf.image.resize_nearest_neighbor(tf.expand_dims(image, 0),
+                                           (height, width))[0]
   png = tf.image.encode_png(image)
   with utils.nogpu_session() as sess:
     res = sess.run(png)

From d94dec305257a0a5c0de8c8933391fb8b8854a3b Mon Sep 17 00:00:00 2001
From: Hermann <blumh@ethz.ch>
Date: Thu, 31 Oct 2019 22:50:18 +0900
Subject: [PATCH 10/11] remove dictionary unpacking

---
 tensorflow_datasets/image/lost_and_found.py | 17 ++++++++---------
 1 file changed, 8 insertions(+), 9 deletions(-)

diff --git a/tensorflow_datasets/image/lost_and_found.py b/tensorflow_datasets/image/lost_and_found.py
index 432447a85c4..173eadd64c6 100644
--- a/tensorflow_datasets/image/lost_and_found.py
+++ b/tensorflow_datasets/image/lost_and_found.py
@@ -125,15 +125,15 @@ def _info(self):
                                            encoding_format='png'),
         'disparity_map': tfds.features.Image(shape=(1024, 2048, 1),
                                              encoding_format='png')}
+    features = {feat: possible_features[feat]
+                for feat in self.builder_config.features}
+    features['image_id'] = tfds.features.Text()
     return tfds.core.DatasetInfo(
         builder=self,
         # This is the description that will appear on the datasets page.
         description=_DESCRIPTION,
         # tfds.features.FeatureConnectors
-        features=tfds.features.FeaturesDict({
-            'image_id': tfds.features.Text(),
-            **{feat: possible_features[feat]
-               for feat in self.builder_config.features}}),
+        features=features,
         # Homepage of the dataset for documentation
         urls=['http://www.6d-vision.com/lostandfounddataset'],
         citation=_CITATION,
@@ -195,11 +195,10 @@ def _generate_examples(self, **paths):
       for left_img in tf.io.gfile.listdir(left_city_root):
         image_id = _get_id_from_left_image(left_img)
 
-        features = {
-            'image_id': image_id,
-            **{feat: path.join(paths_city_root[feat],
-                               '{}_{}.png'.format(image_id, file_suffix[feat]))
-               for feat in paths}}
+        features = {feat: path.join(paths_city_root[feat],
+                                    '{}_{}.png'.format(image_id, file_suffix[feat]))
+                    for feat in paths}
+        features['image_id'] = image_id
 
         yield image_id, features
 

From 599c94fcfb156afe39b457af9fe580d11e41f0b2 Mon Sep 17 00:00:00 2001
From: Hermann <blumh@ethz.ch>
Date: Fri, 1 Nov 2019 12:36:51 +0900
Subject: [PATCH 11/11] fix featuredict

---
 tensorflow_datasets/image/lost_and_found.py | 1 +
 1 file changed, 1 insertion(+)

diff --git a/tensorflow_datasets/image/lost_and_found.py b/tensorflow_datasets/image/lost_and_found.py
index 173eadd64c6..ad5a9d0e5ec 100644
--- a/tensorflow_datasets/image/lost_and_found.py
+++ b/tensorflow_datasets/image/lost_and_found.py
@@ -128,6 +128,7 @@ def _info(self):
     features = {feat: possible_features[feat]
                 for feat in self.builder_config.features}
     features['image_id'] = tfds.features.Text()
+    features = tfds.features.FeaturesDict(features)
     return tfds.core.DatasetInfo(
         builder=self,
         # This is the description that will appear on the datasets page.