R8v4gLDU

1"""Fine-Grained Visual Classification of Aircraft (FGVC-Aircraft) is a benchmark dataset for the fine grained visual categorization of aircraft."""
2
3from __future__ import absolute_import
4from __future__ import division
5from __future__ import print_function
6
7import tensorflow_datasets.public_api as tfds
8
9_CITATION = """@techreport{maji13fine-grained,
10   title = {Fine-Grained Visual Classification of Aircraft},
11   author = {S. Maji and J. Kannala and E. Rahtu and M. Blaschko and A. Vedaldi},
12   year = {2013},
13   archivePrefix = {arXiv},
14   eprint = {1306.5151},
15   primaryClass = "cs-cv",
16}
17"""
18
19_DESCRIPTION = """
20Fine-Grained Visual Classification of Aircraft (FGVC-Aircraft) is a benchmark dataset for the fine grained visual categorization of aircraft.
21"""
22
23_URL = "http://www.robots.ox.ac.uk/~vgg/data/fgvc-aircraft/archives/fgvc-aircraft-2013b.tar.gz" 
24
25class Fgvc-aircraft(tfds.core.GeneratorBasedBuilder):
26
27  VERSION = tfds.core.Version('1.0.0')
28  SUPPORTED_VERSIONS = [
29      tfds.core.Version(
30          "3.0.0", "New split API (https://tensorflow.org/datasets/splits)"),
31  ]
32
33  def _info(self):
34    return tfds.core.DatasetInfo(
35        builder=self,
36        description=_DESCRIPTION,
37        #Are these valid arguments for features?
38        features=tfds.features.FeaturesDict({
39            "image": tfds.features.Image(),
40            "variant": tfds.features.ClassLabel(num_classes=102),
41            "family": tfds.features.ClassLabel(num_classes=70),
42            "manufacturer": tfds.features.ClassLabel(num_classes=41),
43            "bbox": tfds.features.BBoxFeature(),
44        }),
45        #Slightly confused what "supervised keys" are, would the following be correct?
46        supervised_keys=("image", "variant", "family", "manufacturer"),
47        homepage="http://www.robots.ox.ac.uk/~vgg/data/fgvc-aircraft/",
48        citation=_CITATION,
49    )
50
51  def _split_generators(self, dl_manager):
52    path = dl_manager.download_and_extract(_URL)
53
54    #how do I split the data? The names of the train/test pictures are in image_train.txt and image_test.txt files.
55    #this is unlike the example document where they have the images in train/test folders.
56    #how would I add image, variant, family, manufacturer, and bbox in this? The information is provided in different txt files for each property.
57
58    return [
59        tfds.core.SplitGenerator(
60            name=tfds.Split.TRAIN,
61            gen_kwargs={},
62        ),
63        tfds.core.SplitGenerator(
64            name=tfds.Split.TEST,
65            gen_kwargs={},
66        ),
67    ]
68
69  def _generate_examples(self):
70    """Yields examples."""
71    
72    #because the properties are in different txt documents, would I have to process all of them first and then yield info?
73    #how would I structure what to yield?
74
75    yield 'key', {}