Merge pull request #1326 from rickwierenga:master

PiperOrigin-RevId: 292867212

Author: copybara-github

docs/release_notes.md

@@ -18,9 +18,8 @@
 *   Add e-SNLI dataset from the paper
     [e-SNLI](http://papers.nips.cc/paper/8163-e-snli-natural-language-inference-with-natural-language-explanations.pdf).
 *   Add SCAN dataset introduced [here](https://arxiv.org/pdf/1711.00350.pdf).
+*   Add [Imagewang](https://github.com/fastai/imagenette) dataset.
 *   Add DIV2K dataset from the paper
     [DIV2K](http://www.vision.ee.ethz.ch/~timofter/publications/Agustsson-CVPRW-2017.pdf)
 *   Add CFQ (Compositional Freebase Questions) dataset from
     [this paper](https://openreview.net/pdf?id=SygcCnNKwr).
-
-

setup.py

@@ -90,6 +90,7 @@
     'image/imagenet2012_labels.txt',
     'image/imagenet2012_validation_labels.txt',
     'image/imagenette_labels.txt',
+    'image/imagewang_labels.txt',
     'image/inaturalist_labels.txt',
     'image/inaturalist_supercategories.txt',
     'image/open_images_classes_all.txt',

tensorflow_datasets/image/__init__.py

@@ -61,6 +61,7 @@
 from tensorflow_datasets.image.imagenet2012_corrupted import Imagenet2012Corrupted
 from tensorflow_datasets.image.imagenet_resized import ImagenetResized
 from tensorflow_datasets.image.imagenette import Imagenette
+from tensorflow_datasets.image.imagewang import Imagewang
 from tensorflow_datasets.image.inaturalist import INaturalist2017
 from tensorflow_datasets.image.lfw import LFW
 from tensorflow_datasets.image.lost_and_found import LostAndFound

tensorflow_datasets/image/imagewang.py

@@ -0,0 +1,144 @@
+# coding=utf-8
+# Copyright 2020 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.
+
+# -*- coding: utf-8 -*-
+"""Imagewang contains Imagenette and Imagewoof combined."""
+
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+
+import os
+
+import tensorflow.compat.v2 as tf
+import tensorflow_datasets.public_api as tfds
+
+_CITATION = """
+@misc{imagewang,
+  author    = "Jeremy Howard",
+  title     = "Imagewang",
+  url       = "https://github.com/fastai/imagenette/"
+}
+"""
+
+_DESCRIPTION = """\
+Imagewang contains Imagenette and Imagewoof combined
+Image网 (pronounced "Imagewang"; 网 means "net" in Chinese) contains Imagenette
+and Imagewoof combined, but with some twists that make it into a tricky
+semi-supervised unbalanced classification problem:
+
+* The validation set is the same as Imagewoof (i.e. 30% of Imagewoof images);
+  there are no Imagenette images in the validation set (they're all in the
+  training set)
+* Only 10% of Imagewoof images are in the training set!
+* The remaining are in the unsup ("unsupervised") directory, and you can not
+  use their labels in training!
+* It's even hard to type and hard to say!
+
+The dataset comes in three variants:
+  * Full size
+  * 320 px
+  * 160 px
+This dataset consists of the Imagenette dataset {size} variant.
+"""
+
+_DESCRIPTION_SHORT = """\
+Imagewang contains Imagenette and Imagewoof combined.
+"""
+
+_LABELS_FNAME = "image/imagewang_labels.txt"
+_URL_PREFIX = "https://s3.amazonaws.com/fast-ai-imageclas"
+_SIZES = ["full-size", "320px", "160px"]
+
+_SIZE_TO_DIRNAME = {
+    "full-size": "imagewang",
+    "320px": "imagewang-320",
+    "160px": "imagewang-160"
+}
+
+
+class ImagewangConfig(tfds.core.BuilderConfig):
+  """BuilderConfig for Imagewang."""
+
+  def __init__(self, size, **kwargs):
+    super(ImagewangConfig, self).__init__(
+        version=tfds.core.Version("2.0.0"), **kwargs)
+    self.size = size
+
+
+def _make_builder_configs():
+  configs = []
+  for size in _SIZES:
+    configs.append(
+        ImagewangConfig(name=size, size=size, description=_DESCRIPTION_SHORT))
+  return configs
+
+
+class Imagewang(tfds.core.GeneratorBasedBuilder):
+  """Imagewang contains Imagenette and Imagewoof combined."""
+
+  BUILDER_CONFIGS = _make_builder_configs()
+
+  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)
+        }),
+        supervised_keys=("image", "label"),
+        homepage="https://github.com/fastai/imagenette",
+        citation=_CITATION,
+    )
+
+  def _split_generators(self, dl_manager):
+    """Returns SplitGenerators."""
+    size = self.builder_config.size
+    if size in _SIZES:
+      size_str = "" if size == "full-size" else "-" + size[:-2]
+      url = os.path.join(_URL_PREFIX, "imagewang%s.tgz" % size_str)
+      path = dl_manager.download_and_extract(url)
+      train_path = os.path.join(path, _SIZE_TO_DIRNAME[size], "train")
+      val_path = os.path.join(path, _SIZE_TO_DIRNAME[size], "val")
+    else:
+      raise ValueError("size must be one of %s" % _SIZES)
+
+    return [
+        tfds.core.SplitGenerator(
+            name=tfds.Split.TRAIN,
+            gen_kwargs={
+                "datapath": train_path,
+            },
+        ),
+        tfds.core.SplitGenerator(
+            name=tfds.Split.VALIDATION,
+            gen_kwargs={
+                "datapath": val_path,
+            },
+        ),
+    ]
+
+  def _generate_examples(self, datapath):
+    """Yields examples."""
+    for label in tf.io.gfile.listdir(datapath):
+      for fpath in tf.io.gfile.glob(os.path.join(datapath, label, "*.JPEG")):
+        fname = os.path.basename(fpath)
+        record = {
+            "image": fpath,
+            "label": label,
+        }
+        yield fname, record

tensorflow_datasets/image/imagewang_labels.txt

@@ -0,0 +1,20 @@
+n03394916
+n03417042
+n02089973
+n02086240
+n02105641
+n02087394
+n02099601
+n02115641
+n03445777
+n02102040
+n03425413
+n03888257
+n03028079
+n03000684
+n01440764
+n02093754
+n02111889
+n02096294
+n02088364
+n02979186

tensorflow_datasets/image/imagewang_test.py

@@ -0,0 +1,54 @@
+# coding=utf-8
+# Copyright 2020 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 Imagewang."""
+
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+
+from tensorflow_datasets import testing
+from tensorflow_datasets.image import imagewang
+
+
+class ImagewangFullSizeTest(testing.DatasetBuilderTestCase):
+  DATASET_CLASS = imagewang.Imagewang
+  BUILDER_CONFIG_NAMES_TO_TEST = ["full-size"]
+  SPLITS = {
+      "train": 4,
+      "validation": 4,
+  }
+
+
+class Imagewang320Test(testing.DatasetBuilderTestCase):
+  DATASET_CLASS = imagewang.Imagewang
+  BUILDER_CONFIG_NAMES_TO_TEST = ["320px"]
+  SPLITS = {
+      "train": 4,
+      "validation": 4,
+  }
+
+
+class Imagewang160Test(testing.DatasetBuilderTestCase):
+  DATASET_CLASS = imagewang.Imagewang
+  BUILDER_CONFIG_NAMES_TO_TEST = ["160px"]
+  SPLITS = {
+      "train": 4,
+      "validation": 4,
+  }
+
+
+if __name__ == "__main__":
+  testing.test_main()