tanganke commited on
Commit
f5f283f
1 Parent(s): 00b6fc6

Upload folder using huggingface_hub

Browse files
Files changed (4) hide show
  1. README.md +95 -0
  2. data/test.zip +3 -0
  3. data/train.zip +3 -0
  4. dtd.py +94 -0
README.md ADDED
@@ -0,0 +1,95 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ dataset_info:
3
+ features:
4
+ - name: image
5
+ dtype: image
6
+ - name: label
7
+ dtype:
8
+ class_label:
9
+ names:
10
+ '0': banded
11
+ '1': blotchy
12
+ '2': braided
13
+ '3': bubbly
14
+ '4': bumpy
15
+ '5': chequered
16
+ '6': cobwebbed
17
+ '7': cracked
18
+ '8': crosshatched
19
+ '9': crystalline
20
+ '10': dotted
21
+ '11': fibrous
22
+ '12': flecked
23
+ '13': freckled
24
+ '14': frilly
25
+ '15': gauzy
26
+ '16': grid
27
+ '17': grooved
28
+ '18': honeycombed
29
+ '19': interlaced
30
+ '20': knitted
31
+ '21': lacelike
32
+ '22': lined
33
+ '23': marbled
34
+ '24': matted
35
+ '25': meshed
36
+ '26': paisley
37
+ '27': perforated
38
+ '28': pitted
39
+ '29': pleated
40
+ '30': polka-dotted
41
+ '31': porous
42
+ '32': potholed
43
+ '33': scaly
44
+ '34': smeared
45
+ '35': spiralled
46
+ '36': sprinkled
47
+ '37': stained
48
+ '38': stratified
49
+ '39': striped
50
+ '40': studded
51
+ '41': swirly
52
+ '42': veined
53
+ '43': waffled
54
+ '44': woven
55
+ '45': wrinkled
56
+ '46': zigzagged
57
+ splits:
58
+ - name: train
59
+ num_bytes: 448550
60
+ num_examples: 3760
61
+ - name: test
62
+ num_bytes: 220515
63
+ num_examples: 1880
64
+ download_size: 625712354
65
+ dataset_size: 669065
66
+ ---
67
+
68
+ # [DTD: Describable Textures Dataset](https://www.robots.ox.ac.uk/~vgg/data/dtd/)
69
+
70
+ The Describable Textures Dataset (DTD) is an evolving collection of textural images in the wild, annotated with a series of human-centric attributes, inspired by the perceptual properties of textures.
71
+ This data is made available to the computer vision community for research purposes
72
+
73
+ ## Usage
74
+
75
+ ```python
76
+ from datasets import load_dataset
77
+
78
+ dataset = load_dataset('tanganke/dtd')
79
+ ```
80
+
81
+ - **Features:**
82
+ - **Image**: The primary data type, which is a digital image used for classification. The format and dimensions of the images are not specified in this snippet but should be included if available.
83
+ - **Label**: A categorical feature representing the texture or pattern class of each image. The dataset includes 46 classes with descriptive names ranging from 'banded' to 'zigzagged'.
84
+ - **Class Labels**:
85
+ - '0': banded
86
+ - '1': blotchy
87
+ - '2': braided
88
+ - ...
89
+ - '45': wrinkled
90
+ - '46': zigzagged
91
+
92
+ - **Splits**: The dataset is divided into training and test subsets for model evaluation.
93
+ - **Training**: containing 3760 examples with a total size of 448,550 bytes.
94
+ - **Test**: containing 1880 examples with a total size of 220,515 bytes.
95
+
data/test.zip ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:173d40121e4f4ee90e09e5082ab76e124355297b3b72670aec49463115c953bd
3
+ size 177943979
data/train.zip ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:81bc25a565bf38343376b04f55d1fd81c0e03fb5a2005734d3c06d16ef99ade2
3
+ size 447768375
dtd.py ADDED
@@ -0,0 +1,94 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import datasets
2
+ from datasets.data_files import DataFilesDict
3
+ from datasets.packaged_modules.imagefolder.imagefolder import ImageFolder, ImageFolderConfig
4
+
5
+ logger = datasets.logging.get_logger(__name__)
6
+
7
+
8
+ class GTSRB(ImageFolder):
9
+ R"""
10
+ DTD dataset for image classification.
11
+ """
12
+
13
+ BUILDER_CONFIG_CLASS = ImageFolderConfig
14
+ BUILDER_CONFIGS = [
15
+ ImageFolderConfig(
16
+ name="default",
17
+ features=("images", "labels"),
18
+ data_files=DataFilesDict({split: f"data/{split}.zip" for split in ["train", "test"]}),
19
+ )
20
+ ]
21
+
22
+ classnames = [
23
+ "banded",
24
+ "blotchy",
25
+ "braided",
26
+ "bubbly",
27
+ "bumpy",
28
+ "chequered",
29
+ "cobwebbed",
30
+ "cracked",
31
+ "crosshatched",
32
+ "crystalline",
33
+ "dotted",
34
+ "fibrous",
35
+ "flecked",
36
+ "freckled",
37
+ "frilly",
38
+ "gauzy",
39
+ "grid",
40
+ "grooved",
41
+ "honeycombed",
42
+ "interlaced",
43
+ "knitted",
44
+ "lacelike",
45
+ "lined",
46
+ "marbled",
47
+ "matted",
48
+ "meshed",
49
+ "paisley",
50
+ "perforated",
51
+ "pitted",
52
+ "pleated",
53
+ "polka-dotted",
54
+ "porous",
55
+ "potholed",
56
+ "scaly",
57
+ "smeared",
58
+ "spiralled",
59
+ "sprinkled",
60
+ "stained",
61
+ "stratified",
62
+ "striped",
63
+ "studded",
64
+ "swirly",
65
+ "veined",
66
+ "waffled",
67
+ "woven",
68
+ "wrinkled",
69
+ "zigzagged",
70
+ ]
71
+
72
+ clip_templates = [
73
+ lambda c: f"a photo of a {c} texture.",
74
+ lambda c: f"a photo of a {c} pattern.",
75
+ lambda c: f"a photo of a {c} thing.",
76
+ lambda c: f"a photo of a {c} object.",
77
+ lambda c: f"a photo of the {c} texture.",
78
+ lambda c: f"a photo of the {c} pattern.",
79
+ lambda c: f"a photo of the {c} thing.",
80
+ lambda c: f"a photo of the {c} object.",
81
+ ]
82
+
83
+ def _info(self):
84
+ return datasets.DatasetInfo(
85
+ description="DTD dataset for image classification.",
86
+ features=datasets.Features(
87
+ {
88
+ "image": datasets.Image(),
89
+ "label": datasets.ClassLabel(names=self.classnames),
90
+ }
91
+ ),
92
+ supervised_keys=("image", "label"),
93
+ task_templates=[datasets.ImageClassification(image_column="image", label_column="label")],
94
+ )