Create gen_script.py
Browse files- gen_script.py +93 -0
gen_script.py
ADDED
@@ -0,0 +1,93 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from functools import cached_property
|
2 |
+
from pathlib import Path
|
3 |
+
|
4 |
+
import datasets
|
5 |
+
|
6 |
+
_VERSION = "0.1.0"
|
7 |
+
|
8 |
+
_CITATION = """
|
9 |
+
@inproceedings{5539970,
|
10 |
+
title = {SUN database: Large-scale scene recognition from abbey to zoo},
|
11 |
+
author = {Xiao, Jianxiong and Hays, James and Ehinger, Krista A. and Oliva, Aude and Torralba, Antonio},
|
12 |
+
year = 2010,
|
13 |
+
booktitle = {2010 IEEE Computer Society Conference on Computer Vision and Pattern Recognition},
|
14 |
+
volume = {},
|
15 |
+
number = {},
|
16 |
+
pages = {3485--3492},
|
17 |
+
doi = {10.1109/CVPR.2010.5539970},
|
18 |
+
keywords = {Sun;Large-scale systems;Layout;Humans;Image databases;Computer vision;Anthropometry;Bridges;Legged locomotion;Spatial databases}
|
19 |
+
}
|
20 |
+
@article{Xiao2014SUNDE,
|
21 |
+
title = {SUN Database: Exploring a Large Collection of Scene Categories},
|
22 |
+
author = {Jianxiong Xiao and Krista A. Ehinger and James Hays and Antonio Torralba and Aude Oliva},
|
23 |
+
year = 2014,
|
24 |
+
journal = {International Journal of Computer Vision},
|
25 |
+
volume = 119,
|
26 |
+
pages = {3--22},
|
27 |
+
url = {https://api.semanticscholar.org/CorpusID:10224573}
|
28 |
+
}
|
29 |
+
"""
|
30 |
+
|
31 |
+
_DESCRIPTION = """
|
32 |
+
Scene categorization is a fundamental problem in computer vision.
|
33 |
+
However, scene understanding research has been constrained by the limited scope of currently-used databases which do not capture the full variety of scene categories.
|
34 |
+
Whereas standard databases for object categorization contain hundreds of different classes of objects, the largest available dataset of scene categories contains only 15 classes.
|
35 |
+
In this paper we propose the extensive Scene UNderstanding (SUN) database that contains 899 categories and 130,519 images.
|
36 |
+
We use 397 well-sampled categories to evaluate numerous state-of-the-art algorithms for scene recognition and establish new bounds of performance.
|
37 |
+
We measure human scene classification performance on the SUN database and compare this with computational methods.
|
38 |
+
"""
|
39 |
+
|
40 |
+
_HOMEPAGE = "https://vision.princeton.edu/projects/2010/SUN/"
|
41 |
+
|
42 |
+
_LICENSE = ""
|
43 |
+
|
44 |
+
_URL = "http://vision.princeton.edu/projects/2010/SUN/SUN397.tar.gz"
|
45 |
+
|
46 |
+
|
47 |
+
class SUN397(datasets.GeneratorBasedBuilder):
|
48 |
+
DEFAULT_WRITER_BATCH_SIZE = 1000
|
49 |
+
|
50 |
+
@cached_property
|
51 |
+
def archive_path(self):
|
52 |
+
dl_manager = datasets.DownloadManager()
|
53 |
+
return Path(dl_manager.download_and_extract(_URL)) / "SUN397"
|
54 |
+
|
55 |
+
@property
|
56 |
+
def features(self):
|
57 |
+
return datasets.Features(
|
58 |
+
{
|
59 |
+
"image": datasets.Image(mode="RGB"),
|
60 |
+
"label": datasets.ClassLabel(names_file=self.archive_path / "ClassName.txt"),
|
61 |
+
}
|
62 |
+
)
|
63 |
+
|
64 |
+
def _info(self):
|
65 |
+
return datasets.DatasetInfo(
|
66 |
+
features=self.features,
|
67 |
+
supervised_keys=None,
|
68 |
+
description=_DESCRIPTION,
|
69 |
+
homepage=_HOMEPAGE,
|
70 |
+
license=_LICENSE,
|
71 |
+
version=_VERSION,
|
72 |
+
citation=_CITATION,
|
73 |
+
)
|
74 |
+
|
75 |
+
def _split_generators(self, dl_manager: datasets.DownloadManager):
|
76 |
+
images = sorted(list(self.archive_path.rglob("*.jpg")))
|
77 |
+
|
78 |
+
return [
|
79 |
+
datasets.SplitGenerator(
|
80 |
+
name=datasets.Split.TRAIN,
|
81 |
+
gen_kwargs={"images": images},
|
82 |
+
),
|
83 |
+
]
|
84 |
+
|
85 |
+
def _generate_examples(self, images: list[Path]):
|
86 |
+
for i, image in enumerate(images):
|
87 |
+
yield (
|
88 |
+
i,
|
89 |
+
{
|
90 |
+
"image": str(image),
|
91 |
+
"label": f"/{image.relative_to(self.archive_path).parent}",
|
92 |
+
},
|
93 |
+
)
|