Datasets:
Tasks:
Image Classification
Sub-tasks:
multi-class-image-classification
Languages:
English
Size:
100K<n<1M
ArXiv:
License:
try csv parse
Browse files- NIH-Chest-X-ray-dataset.py +29 -3
NIH-Chest-X-ray-dataset.py
CHANGED
@@ -3,6 +3,8 @@ import os
|
|
3 |
import datasets
|
4 |
from datasets.tasks import ImageClassification
|
5 |
|
|
|
|
|
6 |
|
7 |
_HOMEPAGE = "https://nihcc.app.box.com/v/ChestXray-NIHCC"
|
8 |
|
@@ -22,7 +24,7 @@ The NIH Chest X-ray dataset consists of 100,000 de-identified images of chest x-
|
|
22 |
The data is provided by the NIH Clinical Center and is available through the NIH download site: https://nihcc.app.box.com/v/ChestXray-NIHCC
|
23 |
"""
|
24 |
|
25 |
-
|
26 |
'https://nihcc.box.com/shared/static/vfk49d74nhbxq3nqjg0900w5nvkorp5c.gz',
|
27 |
'https://nihcc.box.com/shared/static/i28rlmbvmfjbl8p2n3ril0pptcmcu9d1.gz',
|
28 |
'https://nihcc.box.com/shared/static/f1t00wrtdk94satdfb9olcolqx20z2jp.gz',
|
@@ -37,6 +39,17 @@ _URLS = [
|
|
37 |
'https://nihcc.box.com/shared/static/ioqwiy20ihqwyr8pf4c24eazhh281pbu.gz'
|
38 |
]
|
39 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
40 |
LABEL2IDX = {'No Finding': 0,
|
41 |
'Atelactasis': 1,
|
42 |
'Cardiomegaly': 2,
|
@@ -53,6 +66,9 @@ LABEL2IDX = {'No Finding': 0,
|
|
53 |
'Pleural_Thickening': 13,
|
54 |
'Hernia': 14}
|
55 |
|
|
|
|
|
|
|
56 |
class XChest(datasets.GeneratorBasedBuilder):
|
57 |
"""NIH Image Chest X-ray dataset."""
|
58 |
|
@@ -63,7 +79,7 @@ class XChest(datasets.GeneratorBasedBuilder):
|
|
63 |
{
|
64 |
"image_file_path": datasets.Value("string"),
|
65 |
"image": datasets.Image(),
|
66 |
-
"labels": datasets.features.ClassLabel(names=
|
67 |
}
|
68 |
),
|
69 |
supervised_keys=("image", "labels"),
|
@@ -72,7 +88,17 @@ class XChest(datasets.GeneratorBasedBuilder):
|
|
72 |
task_templates=[ImageClassification(image_column="image",
|
73 |
label_column="labels")],
|
74 |
)
|
75 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
76 |
|
77 |
|
78 |
|
|
|
3 |
import datasets
|
4 |
from datasets.tasks import ImageClassification
|
5 |
|
6 |
+
from requests import get
|
7 |
+
|
8 |
|
9 |
_HOMEPAGE = "https://nihcc.app.box.com/v/ChestXray-NIHCC"
|
10 |
|
|
|
24 |
The data is provided by the NIH Clinical Center and is available through the NIH download site: https://nihcc.app.box.com/v/ChestXray-NIHCC
|
25 |
"""
|
26 |
|
27 |
+
_IMAGE_URLS2 = [
|
28 |
'https://nihcc.box.com/shared/static/vfk49d74nhbxq3nqjg0900w5nvkorp5c.gz',
|
29 |
'https://nihcc.box.com/shared/static/i28rlmbvmfjbl8p2n3ril0pptcmcu9d1.gz',
|
30 |
'https://nihcc.box.com/shared/static/f1t00wrtdk94satdfb9olcolqx20z2jp.gz',
|
|
|
39 |
'https://nihcc.box.com/shared/static/ioqwiy20ihqwyr8pf4c24eazhh281pbu.gz'
|
40 |
]
|
41 |
|
42 |
+
_IMAGE_URLS = [
|
43 |
+
'https://huggingface.co/datasets/alkzar90/NIH-Chest-X-ray-dataset/resolve/main/dummy/0.0.0/images_001.tar.gz',
|
44 |
+
'https://huggingface.co/datasets/alkzar90/NIH-Chest-X-ray-dataset/resolve/main/dummy/0.0.0/images_002.tar.gz'
|
45 |
+
]
|
46 |
+
|
47 |
+
_URLS = {
|
48 |
+
'train_val_list': 'https://huggingface.co/datasets/alkzar90/NIH-Chest-X-ray-dataset/raw/main/dummy/0.0.0/train_val_list.txt',
|
49 |
+
'test_list': 'https://huggingface.co/datasets/alkzar90/NIH-Chest-X-ray-dataset/raw/main/dummy/0.0.0/test_list.txt',
|
50 |
+
'image_urls': _IMAGE_URLS
|
51 |
+
}
|
52 |
+
|
53 |
LABEL2IDX = {'No Finding': 0,
|
54 |
'Atelactasis': 1,
|
55 |
'Cardiomegaly': 2,
|
|
|
66 |
'Pleural_Thickening': 13,
|
67 |
'Hernia': 14}
|
68 |
|
69 |
+
_NAMES = list(LABEL2IDX.keys())
|
70 |
+
|
71 |
+
|
72 |
class XChest(datasets.GeneratorBasedBuilder):
|
73 |
"""NIH Image Chest X-ray dataset."""
|
74 |
|
|
|
79 |
{
|
80 |
"image_file_path": datasets.Value("string"),
|
81 |
"image": datasets.Image(),
|
82 |
+
"labels": datasets.features.ClassLabel(names=_NAMES),
|
83 |
}
|
84 |
),
|
85 |
supervised_keys=("image", "labels"),
|
|
|
88 |
task_templates=[ImageClassification(image_column="image",
|
89 |
label_column="labels")],
|
90 |
)
|
91 |
+
|
92 |
+
|
93 |
+
def _split_generators(self, dl_manager):
|
94 |
+
data_files = dl_manager.download_and_extract(_URLS['image_urls'])
|
95 |
+
train_val_list = get(_URLS['train_val_list']).iter_lines()
|
96 |
+
train_val_list = set([x.decode('UTF8') for x in train_val_list])
|
97 |
+
print(train_val_list)
|
98 |
+
pass
|
99 |
+
|
100 |
+
def _generate_examples(self, files):
|
101 |
+
pass
|
102 |
|
103 |
|
104 |
|