Adds Caltech-101. Includes only the image-level labels, and not the bounding boxes.

PiperOrigin-RevId: 239160178

Author: TensorFlow Datasets Team

tensorflow_datasets/image/__init__.py

@@ -15,6 +15,7 @@
 
 """Image datasets."""
 
+from tensorflow_datasets.image.caltech import Caltech101
 from tensorflow_datasets.image.cats_vs_dogs import CatsVsDogs
 from tensorflow_datasets.image.celeba import CelebA
 from tensorflow_datasets.image.celebahq import CelebAHq

tensorflow_datasets/image/caltech.py

@@ -0,0 +1,102 @@
+# 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.
+
+"""Caltech images dataset."""
+
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+
+import os
+import tensorflow as tf
+import tensorflow_datasets.public_api as tfds
+
+_CITATION = """\
+@article{FeiFei2004LearningGV,
+  title={Learning Generative Visual Models from Few Training Examples: An Incremental Bayesian Approach Tested on 101 Object Categories},
+  author={Li Fei-Fei and Rob Fergus and Pietro Perona},
+  journal={Computer Vision and Pattern Recognition Workshop},
+  year={2004},
+}
+"""
+_DESCRIPTION = """\
+Caltech-101 consists of pictures of objects belonging to 101 classes, plus
+one `background clutter` class. Each image is labelled with a single object.
+Each class contains roughly 40 to 800 images, totalling around 9k images.
+Images are of variable sizes, with typical edge lengths of 200-300 pixels.
+This version contains image-level labels only. The original dataset also
+contains bounding boxes.
+"""
+_LABELS_FNAME = "image/caltech101_labels.txt"
+_URL = "http://www.vision.caltech.edu/Image_Datasets/Caltech101/"
+_IMAGES_FNAME = "101_ObjectCategories.tar.gz"
+
+
+class Caltech101(tfds.core.GeneratorBasedBuilder):
+  """Caltech-101."""
+
+  VERSION = tfds.core.Version("1.0.0")
+
+  def _info(self):
+    names_file = tfds.core.get_tfds_path(_LABELS_FNAME)
+    return tfds.core.DatasetInfo(
+        builder=self,
+        description=_DESCRIPTION,
+        features=tfds.features.FeaturesDict({
+            "image": tfds.features.Image(),
+            "label": tfds.features.ClassLabel(names_file=names_file),
+            "image/file_name": tfds.features.Text(),  # E.g. 'image_0001.jpg'.
+        }),
+        supervised_keys=("image", "label"),
+        urls=[_URL],
+        citation=_CITATION
+        )
+
+  def _split_generators(self, dl_manager):
+    path = dl_manager.download_and_extract(os.path.join(_URL, _IMAGES_FNAME))
+    # There is no predefined train/val/test split for this dataset.
+    return [
+        tfds.core.SplitGenerator(
+            name=tfds.Split.TRAIN,
+            num_shards=5,
+            gen_kwargs={
+                "images_dir_path": path
+            }),
+    ]
+
+  def _generate_examples(self, images_dir_path):
+    """Generates images and labels given the image directory path.
+
+    Args:
+      images_dir_path: path to the directory where the images are stored.
+
+    Yields:
+      The image path, and its corresponding label and filename.
+    """
+    parent_dir = tf.io.gfile.listdir(images_dir_path)[0]
+    walk_dir = os.path.join(images_dir_path, parent_dir)
+    dirs = tf.io.gfile.listdir(walk_dir)
+
+    for d in dirs:
+      if tf.io.gfile.isdir(os.path.join(walk_dir, d)):
+        for full_path, _, fname in tf.io.gfile.walk(os.path.join(walk_dir, d)):
+          for image_file in fname:
+            if image_file.endswith(".jpg"):
+              image_path = os.path.join(full_path, image_file)
+              yield {
+                  "image": image_path,
+                  "label": d.lower(),
+                  "image/file_name": image_file,
+              }

tensorflow_datasets/image/caltech101_labels.txt

@@ -0,0 +1,102 @@
+accordion
+airplanes
+anchor
+ant
+background_google
+barrel
+bass
+beaver
+binocular
+bonsai
+brain
+brontosaurus
+buddha
+butterfly
+camera
+cannon
+car_side
+ceiling_fan
+cellphone
+chair
+chandelier
+cougar_body
+cougar_face
+crab
+crayfish
+crocodile
+crocodile_head
+cup
+dalmatian
+dollar_bill
+dolphin
+dragonfly
+electric_guitar
+elephant
+emu
+euphonium
+ewer
+faces
+faces_easy
+ferry
+flamingo
+flamingo_head
+garfield
+gerenuk
+gramophone
+grand_piano
+hawksbill
+headphone
+hedgehog
+helicopter
+ibis
+inline_skate
+joshua_tree
+kangaroo
+ketch
+lamp
+laptop
+leopards
+llama
+lobster
+lotus
+mandolin
+mayfly
+menorah
+metronome
+minaret
+motorbikes
+nautilus
+octopus
+okapi
+pagoda
+panda
+pigeon
+pizza
+platypus
+pyramid
+revolver
+rhino
+rooster
+saxophone
+schooner
+scissors
+scorpion
+sea_horse
+snoopy
+soccer_ball
+stapler
+starfish
+stegosaurus
+stop_sign
+strawberry
+sunflower
+tick
+trilobite
+umbrella
+watch
+water_lilly
+wheelchair
+wild_cat
+windsor_chair
+wrench
+yin_yang

tensorflow_datasets/image/caltech_test.py

@@ -0,0 +1,33 @@
+# 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.
+
+"""Tests for flowers data loading."""
+
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+from tensorflow_datasets import testing
+from tensorflow_datasets.image import caltech
+
+
+class Caltech101Test(testing.DatasetBuilderTestCase):
+  DATASET_CLASS = caltech.Caltech101
+
+  SPLITS = {
+      'train': 6
+  }
+
+if __name__ == '__main__':
+  testing.test_main()

tensorflow_datasets/testing/caltech.py

@@ -0,0 +1,94 @@
+# 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.
+
+r"""Script to generate Caltech101 like files with random data for testing.
+
+"""
+
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+
+import os
+
+from absl import app
+from absl import flags
+
+import numpy as np
+import tensorflow as tf
+
+from tensorflow_datasets.core import utils
+from tensorflow_datasets.core.utils import py_utils
+from tensorflow_datasets.image import caltech
+import tensorflow_datasets.public_api as tfds
+from tensorflow_datasets.testing import fake_data_utils
+
+flags.DEFINE_string("tfds_dir", py_utils.tfds_dir(),
+                    "Path to tensorflow_datasets directory")
+
+FLAGS = flags.FLAGS
+
+NUM_CLASSES = 3
+IMAGES_PER_CLASS = 2
+MIN_EDGE_LENGTH = 150
+MAX_EDGE_LENGTH = 350
+
+
+def _output_dir():
+  """Returns output directory."""
+  return os.path.join(FLAGS.tfds_dir, "testing", "test_data", "fake_examples",
+                      "caltech101", "{}_ObjectCategories".format(NUM_CLASSES))
+
+
+def _save_image(jpeg, label, image_idx):
+  """Saves jpeg."""
+  dirname = os.path.join(_output_dir(), label)
+  if not os.path.exists(dirname):
+    os.makedirs(dirname)
+  path = os.path.join(dirname, "image_{:04d}.jpg".format(image_idx))
+  with open(path, "wb") as out_file:
+    out_file.write(jpeg)
+
+
+def _get_jpeg(height, width):
+  """Returns jpeg picture."""
+  image = fake_data_utils.get_random_picture(height, width)
+  jpeg = tf.image.encode_jpeg(image)
+  with utils.nogpu_session() as sess:
+    res = sess.run(jpeg)
+  return res
+
+
+def _generate_images():
+  """Generates training images."""
+  names_file = tfds.core.get_tfds_path(caltech._LABELS_FNAME)  # pylint: disable=protected-access
+  label_names = tfds.features.ClassLabel(
+      names_file=names_file).names[:NUM_CLASSES]
+  for label in label_names:
+    for i in range(IMAGES_PER_CLASS):
+      height = np.random.randint(low=MIN_EDGE_LENGTH, high=MAX_EDGE_LENGTH)
+      width = np.random.randint(low=MIN_EDGE_LENGTH, high=MAX_EDGE_LENGTH)
+      jpeg = _get_jpeg(height=height, width=width)
+      _save_image(jpeg, label, i + 1)
+
+
+def main(argv):
+  if len(argv) > 1:
+    raise app.UsageError("Too many command-line arguments.")
+  _generate_images()
+
+
+if __name__ == "__main__":
+  app.run(main)