Create kmnist.py
Browse files
kmnist.py
ADDED
@@ -0,0 +1,84 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import numpy as np
|
2 |
+
import datasets
|
3 |
+
from PIL import Image
|
4 |
+
|
5 |
+
|
6 |
+
class KMNIST(datasets.GeneratorBasedBuilder):
|
7 |
+
"""Kuzushiji-MNIST and Kuzushiji-49 datasets."""
|
8 |
+
VERSION = datasets.Version("1.0.0")
|
9 |
+
|
10 |
+
BUILDER_CONFIGS = [
|
11 |
+
datasets.BuilderConfig(name="kmnist", description="Kuzushiji-MNIST dataset with 10 classes."),
|
12 |
+
datasets.BuilderConfig(name="k49mnist", description="Kuzushiji-49 dataset with 49 classes."),
|
13 |
+
]
|
14 |
+
|
15 |
+
def _info(self):
|
16 |
+
if self.config.name == "kmnist":
|
17 |
+
num_classes = 10
|
18 |
+
else:
|
19 |
+
num_classes = 49
|
20 |
+
return datasets.DatasetInfo(
|
21 |
+
description="Kuzushiji-MNIST and Kuzushiji-49 datasets.",
|
22 |
+
features=datasets.Features({
|
23 |
+
"image": datasets.Image(), # Automatically converts to PIL.Image
|
24 |
+
"label": datasets.ClassLabel(num_classes=num_classes),
|
25 |
+
}),
|
26 |
+
supervised_keys=("image", "label"),
|
27 |
+
license="CC BY-SA 4.0",
|
28 |
+
homepage="https://github.com/rois-codh/kmnist",
|
29 |
+
citation="""
|
30 |
+
@online{clanuwat2018deep,
|
31 |
+
author = {Tarin Clanuwat and Mikel Bober-Irizar and Asanobu Kitamoto and Alex Lamb and Kazuaki Yamamoto and David Ha},
|
32 |
+
title = {Deep Learning for Classical Japanese Literature},
|
33 |
+
date = {2018-12-03},
|
34 |
+
year = {2018},
|
35 |
+
eprintclass = {cs.CV},
|
36 |
+
eprinttype = {arXiv},
|
37 |
+
eprint = {cs.CV/1812.01718},
|
38 |
+
}
|
39 |
+
"""
|
40 |
+
)
|
41 |
+
|
42 |
+
def _split_generators(self, dl_manager):
|
43 |
+
urls = {
|
44 |
+
"kmnist": {
|
45 |
+
"train_imgs": "http://codh.rois.ac.jp/kmnist/dataset/kmnist/kmnist-train-imgs.npz",
|
46 |
+
"train_labels": "http://codh.rois.ac.jp/kmnist/dataset/kmnist/kmnist-train-labels.npz",
|
47 |
+
"test_imgs": "http://codh.rois.ac.jp/kmnist/dataset/kmnist/kmnist-test-imgs.npz",
|
48 |
+
"test_labels": "http://codh.rois.ac.jp/kmnist/dataset/kmnist/kmnist-test-labels.npz",
|
49 |
+
},
|
50 |
+
"k49mnist": {
|
51 |
+
"train_imgs": "http://codh.rois.ac.jp/kmnist/dataset/k49/k49-train-imgs.npz",
|
52 |
+
"train_labels": "http://codh.rois.ac.jp/kmnist/dataset/k49/k49-train-labels.npz",
|
53 |
+
"test_imgs": "http://codh.rois.ac.jp/kmnist/dataset/k49/k49-test-imgs.npz",
|
54 |
+
"test_labels": "http://codh.rois.ac.jp/kmnist/dataset/k49/k49-test-labels.npz",
|
55 |
+
},
|
56 |
+
}
|
57 |
+
selected_urls = urls[self.config.name]
|
58 |
+
downloaded_files = dl_manager.download(selected_urls)
|
59 |
+
|
60 |
+
return [
|
61 |
+
datasets.SplitGenerator(
|
62 |
+
name=datasets.Split.TRAIN,
|
63 |
+
gen_kwargs={
|
64 |
+
"images_path": downloaded_files["train_imgs"],
|
65 |
+
"labels_path": downloaded_files["train_labels"]
|
66 |
+
}
|
67 |
+
),
|
68 |
+
datasets.SplitGenerator(
|
69 |
+
name=datasets.Split.TEST,
|
70 |
+
gen_kwargs={
|
71 |
+
"images_path": downloaded_files["test_imgs"],
|
72 |
+
"labels_path": downloaded_files["test_labels"]
|
73 |
+
}
|
74 |
+
),
|
75 |
+
]
|
76 |
+
|
77 |
+
def _generate_examples(self, images_path, labels_path):
|
78 |
+
images = np.load(images_path)["arr_0"]
|
79 |
+
labels = np.load(labels_path)["arr_0"]
|
80 |
+
|
81 |
+
for idx, (image, label) in enumerate(zip(images, labels)):
|
82 |
+
# Convert each image to a PIL.Image object
|
83 |
+
image = Image.fromarray(image, mode="L") # Mode "L" for grayscale images
|
84 |
+
yield idx, {"image": image, "label": int(label)}
|