download annotation if auth token exists (#11)
Browse files* download annotation if auth token exists
* fix for CI
- COCOA.py +28 -9
- tests/COCOA_test.py +13 -1
COCOA.py
CHANGED
@@ -70,6 +70,10 @@ _URLS = {
|
|
70 |
"BSDS": {
|
71 |
"images": "http://www.eecs.berkeley.edu/Research/Projects/CS/vision/grouping/BSR/BSR_bsds500.tgz",
|
72 |
},
|
|
|
|
|
|
|
|
|
73 |
}
|
74 |
|
75 |
|
@@ -334,7 +338,7 @@ class CocoaDataset(ds.GeneratorBasedBuilder):
|
|
334 |
return ann_json
|
335 |
|
336 |
@property
|
337 |
-
def
|
338 |
return (
|
339 |
"To use COCOA, you need to download the annotations "
|
340 |
"from the google drive in the official repositories "
|
@@ -489,10 +493,11 @@ class CocoaDataset(ds.GeneratorBasedBuilder):
|
|
489 |
),
|
490 |
]
|
491 |
|
492 |
-
def
|
493 |
-
|
494 |
-
|
495 |
|
|
|
496 |
assert dl_manager.manual_dir is not None, dl_manager.manual_dir
|
497 |
data_path = os.path.expanduser(dl_manager.manual_dir)
|
498 |
|
@@ -501,7 +506,7 @@ class CocoaDataset(ds.GeneratorBasedBuilder):
|
|
501 |
f"{data_path} does not exists. Make sure you insert a manual dir "
|
502 |
'via `datasets.load_dataset("shunk031/COCOA", data_dir=...)` '
|
503 |
"that includes tar/untar files from the COCOA annotation tar.gz. "
|
504 |
-
f"Manual download instructions: {self.
|
505 |
)
|
506 |
else:
|
507 |
data_path = (
|
@@ -509,16 +514,30 @@ class CocoaDataset(ds.GeneratorBasedBuilder):
|
|
509 |
if not os.path.isdir(data_path)
|
510 |
else data_path
|
511 |
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
512 |
|
513 |
assert isinstance(data_path, str)
|
514 |
ann_dir = os.path.join(data_path, "annotations")
|
515 |
|
516 |
if self.config.name == "COCO":
|
517 |
-
return self._split_generators_coco(
|
518 |
-
|
|
|
|
|
519 |
elif self.config.name == "BSDS":
|
520 |
-
return self._split_generators_bsds(
|
521 |
-
|
|
|
|
|
522 |
else:
|
523 |
raise ValueError(f"Invalid name: {self.config.name}")
|
524 |
|
|
|
70 |
"BSDS": {
|
71 |
"images": "http://www.eecs.berkeley.edu/Research/Projects/CS/vision/grouping/BSR/BSR_bsds500.tgz",
|
72 |
},
|
73 |
+
# The author of this loading script has uploaded the annotation files to the HuggingFace's private repository to facilitate testing.
|
74 |
+
# If you are using this loading script, please download the annotations from the appropriate channels, such as the Google Drive link provided by the COCOA's author.
|
75 |
+
# (To the author of COCOA, if there are any issues regarding this matter, please contact us. We will address it promptly.)
|
76 |
+
"annotations": "https://huggingface.co/datasets/shunk031/COCOA-annotation/resolve/main/annotations.tar.gz",
|
77 |
}
|
78 |
|
79 |
|
|
|
338 |
return ann_json
|
339 |
|
340 |
@property
|
341 |
+
def _manual_download_instructions(self) -> str:
|
342 |
return (
|
343 |
"To use COCOA, you need to download the annotations "
|
344 |
"from the google drive in the official repositories "
|
|
|
493 |
),
|
494 |
]
|
495 |
|
496 |
+
def _download_annotation_from_hf(self, dl_manager: ds.DownloadManager) -> str:
|
497 |
+
data_path = dl_manager.download_and_extract(_URLS["annotations"])
|
498 |
+
return data_path # type: ignore
|
499 |
|
500 |
+
def _download_annotation_from_local(self, dl_manager: ds.DownloadManager) -> str:
|
501 |
assert dl_manager.manual_dir is not None, dl_manager.manual_dir
|
502 |
data_path = os.path.expanduser(dl_manager.manual_dir)
|
503 |
|
|
|
506 |
f"{data_path} does not exists. Make sure you insert a manual dir "
|
507 |
'via `datasets.load_dataset("shunk031/COCOA", data_dir=...)` '
|
508 |
"that includes tar/untar files from the COCOA annotation tar.gz. "
|
509 |
+
f"Manual download instructions: {self._manual_download_instructions}"
|
510 |
)
|
511 |
else:
|
512 |
data_path = (
|
|
|
514 |
if not os.path.isdir(data_path)
|
515 |
else data_path
|
516 |
)
|
517 |
+
return data_path # type: ignore
|
518 |
+
|
519 |
+
def _split_generators(self, dl_manager: ds.DownloadManager):
|
520 |
+
urls = _URLS[self.config.name]
|
521 |
+
image_dirs = dl_manager.download_and_extract(urls["images"]) # type: ignore
|
522 |
+
|
523 |
+
if dl_manager.download_config.token:
|
524 |
+
data_path = self._download_annotation_from_hf(dl_manager)
|
525 |
+
else:
|
526 |
+
data_path = self._download_annotation_from_local(dl_manager)
|
527 |
|
528 |
assert isinstance(data_path, str)
|
529 |
ann_dir = os.path.join(data_path, "annotations")
|
530 |
|
531 |
if self.config.name == "COCO":
|
532 |
+
return self._split_generators_coco(
|
533 |
+
ann_dir=ann_dir,
|
534 |
+
image_dirs=image_dirs, # type: ignore
|
535 |
+
)
|
536 |
elif self.config.name == "BSDS":
|
537 |
+
return self._split_generators_bsds(
|
538 |
+
ann_dir=ann_dir,
|
539 |
+
image_dir=image_dirs, # type: ignore
|
540 |
+
)
|
541 |
else:
|
542 |
raise ValueError(f"Invalid name: {self.config.name}")
|
543 |
|
tests/COCOA_test.py
CHANGED
@@ -18,6 +18,13 @@ def data_dir() -> str:
|
|
18 |
return "annotations.tar.gz"
|
19 |
|
20 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
21 |
@pytest.mark.parametrize(
|
22 |
argnames="decode_rle,",
|
23 |
argvalues=(False, True),
|
@@ -42,9 +49,14 @@ def test_load_dataset(
|
|
42 |
expected_num_validation: int,
|
43 |
expected_num_test: int,
|
44 |
decode_rle: bool,
|
|
|
45 |
):
|
46 |
dataset = ds.load_dataset(
|
47 |
-
path=dataset_path,
|
|
|
|
|
|
|
|
|
48 |
)
|
49 |
|
50 |
assert dataset["train"].num_rows == expected_num_train # type: ignore
|
|
|
18 |
return "annotations.tar.gz"
|
19 |
|
20 |
|
21 |
+
@pytest.mark.parametrize(
|
22 |
+
argnames="use_auth_token",
|
23 |
+
argvalues=(
|
24 |
+
True,
|
25 |
+
False,
|
26 |
+
),
|
27 |
+
)
|
28 |
@pytest.mark.parametrize(
|
29 |
argnames="decode_rle,",
|
30 |
argvalues=(False, True),
|
|
|
49 |
expected_num_validation: int,
|
50 |
expected_num_test: int,
|
51 |
decode_rle: bool,
|
52 |
+
use_auth_token: bool,
|
53 |
):
|
54 |
dataset = ds.load_dataset(
|
55 |
+
path=dataset_path,
|
56 |
+
name=dataset_name,
|
57 |
+
data_dir=data_dir if not use_auth_token else None,
|
58 |
+
decode_rle=decode_rle,
|
59 |
+
token=use_auth_token,
|
60 |
)
|
61 |
|
62 |
assert dataset["train"].num_rows == expected_num_train # type: ignore
|