def test_can_convert_voc_to_yolo(self):
target_dataset = Dataset.from_iterable(
[
DatasetItem(id='2007_000001',
subset='train',
image=np.ones((10, 20, 3)),
annotations=[
Bbox(1.0, 2.0, 2.0, 2.0, label=8),
Bbox(4.0, 5.0, 2.0, 2.0, label=15),
Bbox(5.5, 6, 2, 2, label=22),
])
],
categories=[
label.name
for label in VOC.make_voc_categories()[AnnotationType.label]
])
with TestDir() as test_dir:
voc_dir = osp.join(
__file__[:__file__.rfind(osp.join('tests', ''))], 'tests',
'assets', 'voc_dataset', 'voc_dataset1')
yolo_dir = osp.join(test_dir, 'yolo_dir')
run(self, 'convert', '-if', 'voc', '-i', voc_dir, '-f', 'yolo',
'-o', yolo_dir, '--', '--save-images')
parsed_dataset = Dataset.import_from(yolo_dir, format='yolo')
compare_datasets(self,
target_dataset,
parsed_dataset,
require_images=True)
def test_can_import_voc_dataset_with_empty_lines_in_subset_lists(self):
expected_dataset = Dataset.from_iterable([
DatasetItem(id='2007_000001', subset='train',
image=np.ones((10, 20, 3)),
annotations=[
Bbox(1.0, 2.0, 2.0, 2.0, label=8, id=1, group=1,
attributes={
'difficult': False,
'truncated': True,
'occluded': False,
'pose': 'Unspecified'
}
)
])
], categories=VOC.make_voc_categories())
rpath = osp.join('ImageSets', 'Main', 'train.txt')
matrix = [
('voc_detection', '', ''),
('voc_detection', 'train', rpath),
]
for format, subset, path in matrix:
with self.subTest(format=format, subset=subset, path=path):
if subset:
expected = expected_dataset.get_subset(subset)
else:
expected = expected_dataset
actual = Dataset.import_from(osp.join(DUMMY_DATASET3_DIR, path),
format)
compare_datasets(self, expected, actual, require_images=True)
def test_can_save_and_load_voc_segmentation_dataset(self):
expected_dataset = Dataset.from_iterable([
DatasetItem(id='2007_000001', subset='train',
image=np.ones((10, 20, 3)),
annotations=[
Mask(image=np.ones([10, 20]), label=2, group=1)
]),
DatasetItem(id='2007_000002', subset='test',
image=np.ones((10, 20, 3))),
], categories=VOC.make_voc_categories())
dataset_dir = osp.join(DUMMY_DATASETS_DIR, 'voc_dataset1')
rpath = osp.join('ImageSets', 'Segmentation', 'train.txt')
matrix = [
('voc_segmentation', '', ''),
('voc_segmentation', 'train', rpath),
('voc', 'train', rpath),
]
for format, subset, path in matrix:
with self.subTest(format=format, subset=subset, path=path):
if subset:
expected = expected_dataset.get_subset(subset)
else:
expected = expected_dataset
with TestDir() as test_dir:
self._test_can_save_and_load(test_dir, dataset_dir,
expected, format, result_path=path, label_map='voc')
def test_can_import_voc_segmentation_dataset(self):
expected_dataset = Dataset.from_iterable([
DatasetItem(id='2007_000001', subset='train',
image=np.ones((10, 20, 3)),
annotations=[
Mask(image=np.ones([10, 20]), label=2, group=1)
]),
DatasetItem(id='2007_000002', subset='test',
image=np.ones((10, 20, 3))),
], categories=VOC.make_voc_categories())
rpath = osp.join('ImageSets', 'Segmentation', 'train.txt')
matrix = [
('voc_segmentation', '', ''),
('voc_segmentation', 'train', rpath),
('voc', 'train', rpath),
]
for format, subset, path in matrix:
with self.subTest(format=format, subset=subset, path=path):
if subset:
expected = expected_dataset.get_subset(subset)
else:
expected = expected_dataset
actual = Dataset.import_from(osp.join(DUMMY_DATASET_DIR, path),
format)
compare_datasets(self, expected, actual, require_images=True)
def test_convert_to_voc_format(self):
"""
<b>Description:</b>
Ensure that the dataset can be converted to VOC format with
command `datum convert`.
<b>Expected results:</b>
A VOC dataset that matches the expected dataset.
<b>Steps:</b>
1. Get path to the source dataset from assets.
2. Convert source dataset to VOC format, using the `convert` command.
3. Verify that resulting dataset is equal to the expected dataset.
"""
label_map = OrderedDict(('label_' + str(i), [None, [], []]) for i in range(10))
label_map['background'] = [None, [], []]
label_map.move_to_end('background', last=False)
expected_dataset = Dataset.from_iterable([
DatasetItem(id='1', subset='default',
image=np.ones((16, 16, 3)),
annotations=[
Bbox(0.0, 4.0, 4.0, 8.0,
attributes={
'difficult': False,
'truncated': False,
'occluded': False,
'visibility': '1.0',
'ignored': 'False'
},
id=1, label=3, group=1
)
]
)
], categories=VOC.make_voc_categories(label_map))
mot_dir = osp.join(__file__[:__file__.rfind(osp.join('tests', ''))],
'tests', 'assets', 'mot_dataset')
with TestDir() as test_dir:
voc_dir = osp.join(test_dir, 'voc')
run(self, 'convert', '-if', 'mot_seq', '-i', mot_dir,
'-f', 'voc', '-o', voc_dir, '--', '--save-images')
target_dataset = Dataset.import_from(voc_dir, format='voc')
compare_datasets(self, expected_dataset, target_dataset,
require_images=True)
def test_can_save_and_load_voc_detection_dataset(self):
expected_dataset = Dataset.from_iterable([
DatasetItem(id='2007_000001', subset='train',
image=np.ones((10, 20, 3)),
annotations=[
Bbox(4.0, 5.0, 2.0, 2.0, label=15, id=2, group=2,
attributes={
'difficult': False,
'truncated': False,
'occluded': False,
**{
a.name : a.value % 2 == 1
for a in VOC.VocAction
}
}
),
Bbox(1.0, 2.0, 2.0, 2.0, label=8, id=1, group=1,
attributes={
'difficult': False,
'truncated': True,
'occluded': False,
'pose': 'Unspecified'
}
)
]),
DatasetItem(id='2007_000002', subset='test',
image=np.ones((10, 20, 3))),
], categories=VOC.make_voc_categories())
dataset_dir = osp.join(DUMMY_DATASETS_DIR, 'voc_dataset1')
rpath = osp.join('ImageSets', 'Main', 'train.txt')
matrix = [
('voc_detection', '', ''),
('voc_detection', 'train', rpath),
]
for format, subset, path in matrix:
with self.subTest(format=format, subset=subset, path=path):
if subset:
expected = expected_dataset.get_subset(subset)
else:
expected = expected_dataset
with TestDir() as test_dir:
self._test_can_save_and_load(test_dir, dataset_dir,
expected, format, result_path=path, label_map='voc')
def test_export_to_voc_format(self):
label_map = OrderedDict(('label_%s' % i, [None, [], []]) for i in range(10))
label_map['background'] = [None, [], []]
label_map.move_to_end('background', last=False)
expected_dataset = Dataset.from_iterable([
DatasetItem(id='1', subset='train', image=np.ones((10, 15, 3)),
annotations=[
Bbox(0.0, 2.0, 4.0, 2.0,
attributes={
'difficult': False,
'truncated': False,
'occluded': False
},
id=1, label=3, group=1
),
Bbox(3.0, 3.0, 2.0, 3.0,
attributes={
'difficult': False,
'truncated': False,
'occluded': False
},
id=2, label=5, group=2
)
]
)
], categories=VOC.make_voc_categories(label_map))
with TestDir() as test_dir:
yolo_dir = osp.join(__file__[:__file__.rfind(osp.join('tests', ''))],
'tests', 'assets', 'yolo_dataset')
run(self, 'create', '-o', test_dir)
run(self, 'import', '-p', test_dir, '-f', 'yolo', yolo_dir)
voc_export = osp.join(test_dir, 'voc_export')
run(self, 'export', '-p', test_dir, '-f', 'voc',
'-o', voc_export, '--', '--save-images')
parsed_dataset = Dataset.import_from(voc_export, format='voc')
compare_datasets(self, expected_dataset, parsed_dataset,
require_images=True)
def test_can_import_voc_layout_dataset(self):
expected_dataset = Dataset.from_iterable([
DatasetItem(id='2007_000001', subset='train',
image=np.ones((10, 20, 3)),
annotations=[
Bbox(4.0, 5.0, 2.0, 2.0, label=15, id=2, group=2,
attributes={
'difficult': False,
'truncated': False,
'occluded': False,
**{
a.name : a.value % 2 == 1
for a in VOC.VocAction
}
}
),
Bbox(5.5, 6.0, 2.0, 2.0, label=22, group=2),
]),
DatasetItem(id='2007_000002', subset='test',
image=np.ones((10, 20, 3))),
], categories=VOC.make_voc_categories())
rpath = osp.join('ImageSets', 'Layout', 'train.txt')
matrix = [
('voc_layout', '', ''),
('voc_layout', 'train', rpath),
('voc', 'train', rpath),
]
for format, subset, path in matrix:
with self.subTest(format=format, subset=subset, path=path):
if subset:
expected = expected_dataset.get_subset(subset)
else:
expected = expected_dataset
actual = Dataset.import_from(osp.join(DUMMY_DATASET_DIR, path),
format)
compare_datasets(self, expected, actual, require_images=True)
def test_label_projection_with_masks(self):
expected_dataset = Dataset.from_iterable([
DatasetItem(id='2007_000001', subset='train',
image=np.ones((10, 20, 3)),
annotations=[
Bbox(1, 2, 2, 2, label=3,
attributes={
'pose': VOC.VocPose(1).name,
'truncated': True,
'difficult': False,
'occluded': False,
},
id=1, group=1,
),
]
),
DatasetItem(id='2007_000002', subset='test',
image=np.ones((10, 20, 3))),
], categories=VOC.make_voc_categories({
'background': [(0, 0, 0), [], []], # Added on export
'a': [(128, 0, 0), [], []], # Generated by the transform
'b': [(0, 128, 0), [], []], # Generated by the transform
'cat': [(64, 0, 0), [], []] # Original
}))
dataset_path = osp.join(DUMMY_DATASETS_DIR, 'voc_dataset1')
with TestDir() as test_dir:
run(self, 'create', '-o', test_dir)
run(self, 'import', '-p', test_dir, '-f', 'voc', dataset_path)
run(self, 'transform', '-p', test_dir,
'-t', 'project_labels', '--', '-l', 'a', '-l', 'b', '-l', 'cat')
parsed_dataset = Dataset.import_from(
osp.join(test_dir, 'source-1'), 'voc')
compare_datasets(self, expected_dataset, parsed_dataset)
def test_can_save_and_load_voc_dataset(self):
source_dataset = Dataset.from_iterable([
DatasetItem(id='2007_000001', subset='train',
image=np.ones((10, 20, 3)),
annotations=[Label(i) for i in range(22) if i % 2 == 1] + [
Bbox(4.0, 5.0, 2.0, 2.0, label=15, id=1, group=1,
attributes={
'difficult': False,
'truncated': False,
'occluded': False,
**{
a.name : a.value % 2 == 1
for a in VOC.VocAction
}
},
),
Bbox(1.0, 2.0, 2.0, 2.0, label=8, id=2, group=2,
attributes={
'difficult': False,
'truncated': True,
'occluded': False,
'pose': 'Unspecified'
}
),
Bbox(5.5, 6.0, 2.0, 2.0, label=22, id=0, group=1),
Mask(image=np.ones([10, 20]), label=2, group=1),
]),
DatasetItem(id='2007_000002', subset='test',
image=np.ones((10, 20, 3)))
], categories=VOC.make_voc_categories())
voc_dir = osp.join(DUMMY_DATASETS_DIR, 'voc_dataset1')
with TestDir() as test_dir:
self._test_can_save_and_load(test_dir, voc_dir, source_dataset,
'voc', label_map='voc')
def categories(self):
return VOC.make_voc_categories()
def categories(self):
return VOC.make_voc_categories(dst_label_map)
def categories(self):
label_map = OrderedDict()
label_map['background'] = [(0, 0, 0), [], []]
label_map['label_1'] = [(1, 2, 3), [], []]
label_map['label_2'] = [(3, 2, 1), [], []]
return VOC.make_voc_categories(label_map)
def categories(self):
label_map = OrderedDict()
label_map['background'] = [None, [], []]
label_map['Label_1'] = [None, [], []]
label_map['label_2'] = [None, [], []]
return VOC.make_voc_categories(label_map)
def test_preparing_dataset_for_train_model(self):
"""
<b>Description:</b>
Testing a particular example of working with VOC dataset.
<b>Expected results:</b>
A VOC dataset that matches the expected result.
<b>Steps:</b>
1. Get path to the source dataset from assets.
2. Create a datumaro project and add source dataset to it.
3. Leave only non-occluded annotations with `filter` command.
4. Split the dataset into subsets with `transform` command.
5. Export the project to a VOC dataset with `export` command.
6. Verify that the resulting dataset is equal to the expected result.
"""
expected_dataset = Dataset.from_iterable([
DatasetItem(id='c', subset='train',
annotations=[
Bbox(3.0, 1.0, 8.0, 5.0,
attributes={
'truncated': False,
'occluded': False,
'difficult': False
},
id=1, label=2, group=1
)
]
),
DatasetItem(id='d', subset='test',
annotations=[
Bbox(4.0, 4.0, 4.0, 4.0,
attributes={
'truncated': False,
'occluded': False,
'difficult': False
},
id=1, label=3, group=1
)
]
)
], categories=VOC.make_voc_categories())
dataset_path = osp.join(DUMMY_DATASETS_DIR, 'voc_dataset2')
with TestDir() as test_dir:
run(self, 'create', '-o', test_dir)
run(self, 'import', '-p', test_dir, '-f', 'voc', dataset_path)
run(self, 'filter', '-p', test_dir, '-m', 'i+a',
'-e', "/item/annotation[occluded='False']")
run(self, 'transform', '-p', test_dir,
'-t', 'random_split', '--', '-s', 'test:.5',
'-s', 'train:.5', '--seed', '1')
export_path = osp.join(test_dir, 'dataset')
run(self, 'export', '-p', test_dir, '-f', 'voc',
'-o', export_path, '--', '--label-map', 'voc')
parsed_dataset = Dataset.import_from(export_path, format='voc')
compare_datasets(self, expected_dataset, parsed_dataset)
def categories(self):
label_map = VOC.make_voc_label_map()
label_map['non_voc_label'] = [None, [], []]
for label_desc in label_map.values():
label_desc[0] = None # rebuild colormap
return VOC.make_voc_categories(label_map)
