Create dtd.py
Browse files
dtd.py
ADDED
@@ -0,0 +1,90 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import tarfile
|
2 |
+
from io import BytesIO
|
3 |
+
from PIL import Image
|
4 |
+
from tqdm import tqdm
|
5 |
+
import datasets
|
6 |
+
|
7 |
+
|
8 |
+
class DTD(datasets.GeneratorBasedBuilder):
|
9 |
+
"""Describable Textures Dataset (DTD)
|
10 |
+
|
11 |
+
DTD is a texture database, consisting of 5640 images, organized according to a list of 47 terms (categories)
|
12 |
+
inspired from human perception. There are 120 images for each category. Image sizes range between 300x300 and
|
13 |
+
640x640, and the images contain at least 90% of the surface representing the category attribute. The images were
|
14 |
+
collected from Google and Flickr by entering our proposed attributes and related terms as search queries. The images
|
15 |
+
were annotated using Amazon Mechanical Turk in several iterations. For each image we provide key attribute (main
|
16 |
+
category) and a list of joint attributes.
|
17 |
+
|
18 |
+
The data is split in three equal parts, in train, validation and test, 40 images per class, for each split. We
|
19 |
+
provide the ground truth annotation for both key and joint attributes, as well as the 10 splits of the data we used
|
20 |
+
for evaluation.
|
21 |
+
"""
|
22 |
+
|
23 |
+
VERSION = datasets.Version("1.0.0")
|
24 |
+
|
25 |
+
def _info(self):
|
26 |
+
return datasets.DatasetInfo(
|
27 |
+
description="""Describing Textures in the Wild (DTD) is a dataset for texture classification.
|
28 |
+
It contains 5640 images organized into 47 categories.""",
|
29 |
+
features=datasets.Features(
|
30 |
+
{
|
31 |
+
"image": datasets.Image(),
|
32 |
+
"label": datasets.ClassLabel(names=[
|
33 |
+
"banded", "blotchy", "braided", "bubbly", "bumpy", "chequered", "cobwebbed",
|
34 |
+
"cracked", "crosshatched", "crystalline", "dotted", "fibrous", "flecked",
|
35 |
+
"freckled", "frilly", "gauzy", "grid", "grooved", "honeycombed", "interlaced",
|
36 |
+
"knitted", "lacelike", "lined", "marbled", "matted", "meshed", "paisley",
|
37 |
+
"perforated", "pitted", "pleated", "polka-dotted", "porous", "potholed", "scaly",
|
38 |
+
"smeared", "spiralled", "sprinkled", "stained", "stratified", "striped",
|
39 |
+
"studded", "swirly", "veined", "waffled", "woven", "wrinkled", "zigzagged"
|
40 |
+
])
|
41 |
+
}
|
42 |
+
),
|
43 |
+
supervised_keys=("image", "label"),
|
44 |
+
homepage="https://www.robots.ox.ac.uk/~vgg/data/dtd/",
|
45 |
+
citation="""@InProceedings{cimpoi14describing,
|
46 |
+
Author = {M. Cimpoi and S. Maji and I. Kokkinos and S. Mohamed and and A. Vedaldi},
|
47 |
+
Title = {Describing Textures in the Wild},
|
48 |
+
Booktitle = {Proceedings of the {IEEE} Conf. on Computer Vision and Pattern Recognition ({CVPR})},
|
49 |
+
Year = {2014}}""",
|
50 |
+
)
|
51 |
+
|
52 |
+
def _split_generators(self, dl_manager):
|
53 |
+
archive_path = dl_manager.download(
|
54 |
+
"https://www.robots.ox.ac.uk/~vgg/data/dtd/download/dtd-r1.0.1.tar.gz"
|
55 |
+
)
|
56 |
+
|
57 |
+
return [
|
58 |
+
datasets.SplitGenerator(
|
59 |
+
name=datasets.Split.TRAIN,
|
60 |
+
gen_kwargs={"archive_path": archive_path, "split": "train"},
|
61 |
+
),
|
62 |
+
datasets.SplitGenerator(
|
63 |
+
name=datasets.Split.VALIDATION,
|
64 |
+
gen_kwargs={"archive_path": archive_path, "split": "val"},
|
65 |
+
),
|
66 |
+
datasets.SplitGenerator(
|
67 |
+
name=datasets.Split.TEST,
|
68 |
+
gen_kwargs={"archive_path": archive_path, "split": "test"},
|
69 |
+
),
|
70 |
+
]
|
71 |
+
|
72 |
+
def _generate_examples(self, archive_path, split):
|
73 |
+
with tarfile.open(archive_path, "r:gz") as tar:
|
74 |
+
split_file = f"dtd/labels/{split}1.txt"
|
75 |
+
file_names = self._read_split_file(tar, split_file)
|
76 |
+
|
77 |
+
for idx, file_name in enumerate(tqdm(file_names, desc=f"Processing {split} split")):
|
78 |
+
member = tar.getmember(f"dtd/images/{file_name}")
|
79 |
+
file = tar.extractfile(member)
|
80 |
+
image = Image.open(BytesIO(file.read())).convert("RGB")
|
81 |
+
|
82 |
+
yield idx, {
|
83 |
+
"image": image,
|
84 |
+
"label": file_name.split("/")[0],
|
85 |
+
}
|
86 |
+
|
87 |
+
def _read_split_file(self, tar, split_file):
|
88 |
+
"""Helper function to read split file from the tar archive."""
|
89 |
+
split_content = tar.extractfile(split_file).read().decode("utf-8")
|
90 |
+
return split_content.splitlines()
|