def test_polygons_to_masks(self):
source_dataset = Dataset.from_iterable([
DatasetItem(id=1, image=np.zeros((5, 10, 3)), annotations=[
Polygon([0, 0, 4, 0, 4, 4]),
Polygon([5, 0, 9, 0, 5, 5]),
]),
])
target_dataset = Dataset.from_iterable([
DatasetItem(id=1, image=np.zeros((5, 10, 3)), annotations=[
Mask(np.array([
[0, 0, 0, 0, 0, 1, 1, 1, 1, 0],
[0, 0, 0, 0, 0, 1, 1, 1, 0, 0],
[0, 0, 0, 0, 0, 1, 1, 0, 0, 0],
[0, 0, 0, 0, 0, 1, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
]),
),
Mask(np.array([
[0, 1, 1, 1, 0, 0, 0, 0, 0, 0],
[0, 0, 1, 1, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 1, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
]),
),
]),
])
actual = transforms.PolygonsToMasks(source_dataset)
compare_datasets(self, target_dataset, actual)
def test_crop_covered_segments(self):
source_dataset = Dataset.from_iterable([
DatasetItem(id=1, image=np.zeros((5, 5, 3)), annotations=[
# The mask is partially covered by the polygon
Mask(np.array([
[0, 0, 1, 1, 1],
[0, 0, 1, 1, 1],
[1, 1, 1, 1, 1],
[1, 1, 1, 0, 0],
[1, 1, 1, 0, 0]],
), z_order=0),
Polygon([1, 1, 4, 1, 4, 4, 1, 4], z_order=1),
]),
])
target_dataset = Dataset.from_iterable([
DatasetItem(id=1, image=np.zeros((5, 5, 3)), annotations=[
Mask(np.array([
[0, 0, 1, 1, 1],
[0, 0, 0, 0, 1],
[1, 0, 0, 0, 1],
[1, 0, 0, 0, 0],
[1, 1, 1, 0, 0]],
), z_order=0),
Polygon([1, 1, 4, 1, 4, 4, 1, 4], z_order=1),
]),
])
actual = transforms.CropCoveredSegments(source_dataset)
compare_datasets(self, target_dataset, actual)
def test_can_import_v2_0_panoptic_with_keeping_category_ids(self):
labels = [f'class-{i}' for i in range(101)]
labels[1] = ('animal--bird', 'animal')
labels[10] = ('construction--barrier--separator', 'construction')
labels[100] = ('object--vehicle--bicycle', 'object')
label_cat = LabelCategories.from_iterable(labels)
mask_cat = MaskCategories({
1: (165, 42, 42),
10: (128, 128, 128),
100: (119, 11, 32)
})
expected_dataset = Dataset.from_iterable([
DatasetItem(id='0', subset='val', annotations=[
Mask(image=np.array([[1, 1, 1, 0, 0]] * 5), id=1, group=1,
label=1, attributes={'is_crowd': True}),
Mask(image=np.array([[0, 0, 0, 1, 1]] * 5), id=2, group=2,
label=10, attributes={'is_crowd': False}),
Polygon(points=[0, 0, 1, 0, 2, 0, 2, 4, 0, 4], label=1),
Polygon(points=[3, 0, 4, 0, 4, 1, 4, 4, 3, 4], label=10),
], image=np.ones((5, 5, 3))),
DatasetItem(id='1', subset='val', annotations=[
Mask(image=np.array([[1, 1, 0, 0, 0]] * 5), id=1, group=1,
label=100, attributes={'is_crowd': False}),
Mask(image=np.array([[0, 0, 1, 0, 0]] * 5), id=2, group=2,
label=10, attributes={'is_crowd': False}),
Mask(image=np.array([[0, 0, 0, 1, 1]] * 5), id=3, group=3,
label=100, attributes={'is_crowd': True}),
Polygon(points=[2, 0, 2, 1, 2, 2, 2, 3, 2, 4], label=10),
Polygon(points=[0, 0, 1, 0, 1, 4, 4, 0, 0, 0], label=100),
Polygon(points=[3, 0, 4, 0, 4, 4, 3, 4, 3, 0], label=100),
], image=np.ones((5, 5, 3))),
DatasetItem(id='2', subset='train', annotations=[
Mask(image=np.array([[1, 0, 0, 0, 0]] * 5), id=1, group=1,
label=1, attributes={'is_crowd': False}),
Mask(image=np.array([[0, 1, 0, 0, 0]] * 5), id=2, group=2,
label=10, attributes={'is_crowd': False}),
Mask(image=np.array([[0, 0, 1, 0, 0]] * 5), id=3, group=3,
label=1, attributes={'is_crowd': False}),
Mask(image=np.array([[0, 0, 0, 1, 0]] * 5), id=4, group=4,
label=10, attributes={'is_crowd': False}),
Mask(image=np.array([[0, 0, 0, 0, 1]] * 5), id=5, group=5,
label=1, attributes={'is_crowd': False}),
Polygon(points=[0, 0, 0, 1, 0, 2, 0, 3, 0, 4], label=1),
Polygon(points=[2, 0, 2, 1, 2, 2, 2, 3, 2, 4], label=1),
Polygon(points=[4, 0, 4, 1, 4, 2, 4, 3, 4, 4], label=1),
Polygon(points=[1, 0, 1, 1, 1, 2, 1, 3, 1, 4], label=10),
Polygon(points=[3, 0, 3, 1, 3, 2, 3, 3, 3, 4], label=10),
], image=np.ones((5, 5, 3))),
], categories={
AnnotationType.label: label_cat,
AnnotationType.mask: mask_cat
})
imported_dataset = Dataset.import_from(DUMMY_DATASET_V2_0,
'mapillary_vistas_panoptic', keep_original_category_ids=True)
compare_datasets(self, expected_dataset, imported_dataset,
require_images=True)
def test_can_save_and_load_bboxes(self):
expected_dataset = Dataset.from_iterable([
DatasetItem(id='a/b/1',
subset='train',
image=np.ones((10, 15, 3)),
annotations=[
Bbox(1, 3, 6, 10),
Bbox(0, 1, 3, 5, attributes={'text': 'word 0'}),
]),
DatasetItem(id=2,
subset='train',
image=np.ones((10, 15, 3)),
annotations=[
Polygon([0, 0, 3, 0, 4, 7, 1, 8],
attributes={'text': 'word 1'}),
Polygon([1, 2, 5, 3, 6, 8, 0, 7]),
]),
DatasetItem(id=3,
subset='train',
image=np.ones((10, 15, 3)),
annotations=[
Polygon([2, 2, 8, 3, 7, 10, 2, 9],
attributes={'text': 'word_2'}),
Bbox(0, 2, 5, 9, attributes={'text': 'word_3'}),
]),
])
with TestDir() as test_dir:
self._test_save_and_load(
expected_dataset,
partial(IcdarTextLocalizationConverter.convert,
save_images=True), test_dir, 'icdar_text_localization')
def test_shapes_to_boxes(self):
source_dataset = Dataset.from_iterable([
DatasetItem(id=1, image=np.zeros((5, 5, 3)),
annotations=[
Mask(np.array([
[0, 0, 1, 1, 1],
[0, 0, 0, 0, 1],
[1, 0, 0, 0, 1],
[1, 0, 0, 0, 0],
[1, 1, 1, 0, 0]],
), id=1),
Polygon([1, 1, 4, 1, 4, 4, 1, 4], id=2),
PolyLine([1, 1, 2, 1, 2, 2, 1, 2], id=3),
Points([2, 2, 4, 2, 4, 4, 2, 4], id=4),
]
),
])
target_dataset = Dataset.from_iterable([
DatasetItem(id=1, image=np.zeros((5, 5, 3)),
annotations=[
Bbox(0, 0, 4, 4, id=1),
Bbox(1, 1, 3, 3, id=2),
Bbox(1, 1, 1, 1, id=3),
Bbox(2, 2, 2, 2, id=4),
]
),
])
actual = transforms.ShapesToBoxes(source_dataset)
compare_datasets(self, target_dataset, actual)
def _load_polygons(self, items):
polygons_dir = osp.join(self._annotations_dir,
MapillaryVistasPath.POLYGON_DIR)
for item_path in glob.glob(osp.join(polygons_dir, '**', '*.json'),
recursive=True):
item_id = osp.splitext(osp.relpath(item_path, polygons_dir))[0]
item = items.get(item_id)
item_info = {}
item_info = parse_json_file(item_path)
image_size = self._get_image_size(item_info)
if image_size and item.has_image:
item.image = Image(path=item.image.path, size=image_size)
polygons = item_info['objects']
annotations = []
for polygon in polygons:
label = polygon['label']
label_id = self._categories[AnnotationType.label].find(label)[0]
if label_id is None:
label_id = self._categories[AnnotationType.label].add(label)
points = [coord for point in polygon['polygon'] for coord in point]
annotations.append(Polygon(label=label_id, points=points))
if item is None:
items[item_id] = DatasetItem(id=item_id, subset=self._subset,
annotations=annotations)
else:
item.annotations.extend(annotations)
def test_can_import_with_meta_file(self):
expected_dataset = Dataset.from_iterable(
[
DatasetItem(id='street/1', subset='training',
image=np.ones((5, 5, 3)),
annotations=[
Polygon([1, 0, 1, 1, 1, 2, 1, 3, 1, 4],
group=1, z_order=0, id=1, label=1,
attributes={'walkin': True}),
Mask(image=np.array([[0, 0, 1, 1, 1]] * 5), label=0,
group=401, z_order=0, id=401),
Mask(image=np.array([[0, 1, 0, 0, 0]] * 5), label=1,
group=1831, z_order=0, id=1831),
Mask(image=np.array([[0, 0, 0, 1, 1]] * 5), label=2,
id=774, group=774, z_order=1),
Mask(image=np.array([[0, 0, 1, 1, 1]] * 5), label=0,
group=0, z_order=0, id=0),
Mask(image=np.array([[0, 1, 0, 0, 0]] * 5), label=1,
group=1, z_order=0, id=1,
attributes={'walkin': True}),
Mask(image=np.array([[0, 0, 0, 1, 1]] * 5), label=2,
group=2, z_order=1, id=2),
])
], categories={AnnotationType.label: LabelCategories.from_iterable([
'car', 'person', 'door', 'rim'])
}
)
imported_dataset = Dataset.import_from(DUMMY_DATASET_DIR_META_FILE, 'ade20k2020')
compare_datasets(self, expected_dataset, imported_dataset,
require_images=True)
def test_can_load_image(self):
expected_dataset = Dataset.from_iterable([
DatasetItem(id='img0', subset='train',
image=np.ones((8, 8, 3)),
annotations=[
Bbox(0, 2, 4, 2, label=0, z_order=1,
attributes={
'occluded': True,
'a1': True, 'a2': 'v3', 'a3': '0003', 'a4': 2.4,
}),
PolyLine([1, 2, 3, 4, 5, 6, 7, 8],
attributes={'occluded': False}),
], attributes={'frame': 0}),
DatasetItem(id='img1', subset='train',
image=np.ones((10, 10, 3)),
annotations=[
Polygon([1, 2, 3, 4, 6, 5], z_order=1,
attributes={'occluded': False}),
Points([1, 2, 3, 4, 5, 6], label=1, z_order=2,
attributes={'occluded': False}),
], attributes={'frame': 1}),
], categories={
AnnotationType.label: LabelCategories.from_iterable([
['label1', '', {'a1', 'a2', 'a3', 'a4'}],
['label2'],
])
})
parsed_dataset = Dataset.import_from(DUMMY_IMAGE_DATASET_DIR, 'cvat')
compare_datasets(self, expected_dataset, parsed_dataset)
def test_transform_to_labels(self):
src_dataset = Dataset.from_iterable([
DatasetItem(id=1, annotations=[
Label(1),
Bbox(1, 2, 3, 4, label=2),
Bbox(1, 3, 3, 3),
Mask(image=np.array([1]), label=3),
Polygon([1, 1, 2, 2, 3, 4], label=4),
PolyLine([1, 3, 4, 2, 5, 6], label=5)
])
], categories=['label%s' % i for i in range(6)])
dst_dataset = Dataset.from_iterable([
DatasetItem(id=1, annotations=[
Label(1),
Label(2),
Label(3),
Label(4),
Label(5)
]),
], categories=['label%s' % i for i in range(6)])
actual = transforms.AnnsToLabels(src_dataset)
compare_datasets(self, dst_dataset, actual)
def test_can_import_v2_0_instances(self):
label_cat = LabelCategories.from_iterable(['animal--bird',
'construction--barrier--separator', 'object--vehicle--bicycle'])
mask_cat = MaskCategories({
0: (165, 42, 42),
1: (128, 128, 128),
2: (119, 11, 32)
})
expected_dataset = Dataset.from_iterable([
DatasetItem(id='0', subset='val', annotations=[
Mask(image=np.array([[1, 1, 1, 0, 0]] * 5), id=0, label=0),
Mask(image=np.array([[0, 0, 0, 1, 1]] * 5), id=0, label=1),
Polygon(points=[0, 0, 1, 0, 2, 0, 2, 4, 0, 4], label=0),
Polygon(points=[3, 0, 4, 0, 4, 1, 4, 4, 3, 4], label=1),
], image=np.ones((5, 5, 3))),
DatasetItem(id='1', subset='val', annotations=[
Mask(image=np.array([[0, 0, 1, 0, 0]] * 5), id=0, label=1),
Mask(image=np.array([[1, 1, 0, 0, 0]] * 5), id=0, label=2),
Mask(image=np.array([[0, 0, 0, 1, 1]] * 5), id=1, label=2),
Polygon(points=[2, 0, 2, 1, 2, 2, 2, 3, 2, 4], label=1),
Polygon(points=[0, 0, 1, 0, 1, 4, 4, 0, 0, 0], label=2),
Polygon(points=[3, 0, 4, 0, 4, 4, 3, 4, 3, 0], label=2),
], image=np.ones((5, 5, 3))),
DatasetItem(id='2', subset='train', annotations=[
Mask(image=np.array([[1, 0, 0, 0, 0]] * 5), id=0, label=0),
Mask(image=np.array([[0, 0, 1, 0, 0]] * 5), id=1, label=0),
Mask(image=np.array([[0, 0, 0, 0, 1]] * 5), id=2, label=0),
Mask(image=np.array([[0, 1, 0, 0, 0]] * 5), id=0, label=1),
Mask(image=np.array([[0, 0, 0, 1, 0]] * 5), id=1, label=1),
Polygon(points=[0, 0, 0, 1, 0, 2, 0, 3, 0, 4], label=0),
Polygon(points=[2, 0, 2, 1, 2, 2, 2, 3, 2, 4], label=0),
Polygon(points=[4, 0, 4, 1, 4, 2, 4, 3, 4, 4], label=0),
Polygon(points=[1, 0, 1, 1, 1, 2, 1, 3, 1, 4], label=1),
Polygon(points=[3, 0, 3, 1, 3, 2, 3, 3, 3, 4], label=1),
], image=np.ones((5, 5, 3))),
], categories={
AnnotationType.label: label_cat,
AnnotationType.mask: mask_cat
})
imported_dataset = Dataset.import_from(DUMMY_DATASET_V2_0,
'mapillary_vistas_instances')
compare_datasets(self, expected_dataset, imported_dataset,
require_images=True)
def test_merge_instance_segments(self):
source_dataset = Dataset.from_iterable([
DatasetItem(id=1, image=np.zeros((5, 5, 3)),
annotations=[
Mask(np.array([
[0, 0, 1, 1, 1],
[0, 0, 0, 0, 1],
[1, 0, 0, 0, 1],
[1, 0, 0, 0, 0],
[1, 1, 1, 0, 0]],
),
z_order=0, group=1),
Polygon([1, 1, 4, 1, 4, 4, 1, 4],
z_order=1, group=1),
Polygon([0, 0, 0, 2, 2, 2, 2, 0],
z_order=1),
]
),
])
target_dataset = Dataset.from_iterable([
DatasetItem(id=1, image=np.zeros((5, 5, 3)),
annotations=[
Mask(np.array([
[0, 0, 1, 1, 1],
[0, 1, 1, 1, 1],
[1, 1, 1, 1, 1],
[1, 1, 1, 1, 0],
[1, 1, 1, 0, 0]],
),
z_order=0, group=1),
Mask(np.array([
[1, 1, 0, 0, 0],
[1, 1, 0, 0, 0],
[0, 0, 0, 0, 0],
[0, 0, 0, 0, 0],
[0, 0, 0, 0, 0]],
),
z_order=1),
]
),
])
actual = transforms.MergeInstanceSegments(source_dataset,
include_polygons=True)
compare_datasets(self, target_dataset, actual)
def test_can_import_v2_0_panoptic_wo_images(self):
label_cat = LabelCategories.from_iterable([
('animal--bird', 'animal'),
('construction--barrier--separator', 'construction'),
('object--vehicle--bicycle', 'object')
])
mask_cat = MaskCategories({
0: (165, 42, 42),
1: (128, 128, 128),
2: (119, 11, 32)
})
expected_dataset = Dataset.from_iterable([
DatasetItem(id='2', subset='dataset', annotations=[
Mask(image=np.array([[1, 0, 0, 0, 0]] * 5), id=1, group=1,
label=0, attributes={'is_crowd': False}),
Mask(image=np.array([[0, 1, 0, 0, 0]] * 5), id=2, group=2,
label=1, attributes={'is_crowd': False}),
Mask(image=np.array([[0, 0, 1, 0, 0]] * 5), id=3, group=3,
label=0, attributes={'is_crowd': False}),
Mask(image=np.array([[0, 0, 0, 1, 0]] * 5), id=4, group=4,
label=1, attributes={'is_crowd': False}),
Mask(image=np.array([[0, 0, 0, 0, 1]] * 5), id=5, group=5,
label=0, attributes={'is_crowd': False}),
Polygon(points=[0, 0, 0, 1, 0, 2, 0, 3, 0, 4], label=0),
Polygon(points=[2, 0, 2, 1, 2, 2, 2, 3, 2, 4], label=0),
Polygon(points=[4, 0, 4, 1, 4, 2, 4, 3, 4, 4], label=0),
Polygon(points=[1, 0, 1, 1, 1, 2, 1, 3, 1, 4], label=1),
Polygon(points=[3, 0, 3, 1, 3, 2, 3, 3, 3, 4], label=1),
])
], categories={
AnnotationType.label: label_cat,
AnnotationType.mask: mask_cat
})
with TestDir() as test_dir:
dataset_path = osp.join(test_dir, 'dataset')
shutil.copytree(osp.join(DUMMY_DATASET_V2_0, 'train'), dataset_path)
shutil.rmtree(osp.join(dataset_path, 'images'))
imported_dataset = Dataset.import_from(dataset_path,
'mapillary_vistas_panoptic')
compare_datasets(self, expected_dataset, imported_dataset,
require_images=True)
def test_can_resize(self):
small_dataset = Dataset.from_iterable([
DatasetItem(id=i, image=np.ones((4, 4)) * i, annotations=[
Label(1),
Bbox(1, 1, 2, 2, label=2),
Polygon([1, 1, 1, 2, 2, 2, 2, 1], label=1),
PolyLine([1, 1, 1, 2, 2, 2, 2, 1], label=2),
Points([1, 1, 1, 2, 2, 2, 2, 1], label=2),
Mask(np.array([
[0, 0, 1, 1],
[1, 0, 0, 1],
[0, 1, 1, 0],
[1, 1, 0, 0],
]))
]) for i in range(3)
], categories=['a', 'b', 'c'])
big_dataset = Dataset.from_iterable([
DatasetItem(id=i, image=np.ones((8, 8)) * i, annotations=[
Label(1),
Bbox(2, 2, 4, 4, label=2),
Polygon([2, 2, 2, 4, 4, 4, 4, 2], label=1),
PolyLine([2, 2, 2, 4, 4, 4, 4, 2], label=2),
Points([2, 2, 2, 4, 4, 4, 4, 2], label=2),
Mask(np.array([
[0, 0, 0, 0, 1, 1, 1, 1],
[0, 0, 0, 0, 1, 1, 1, 1],
[1, 1, 0, 0, 0, 0, 1, 1],
[1, 1, 0, 0, 0, 0, 1, 1],
[0, 0, 1, 1, 1, 1, 0, 0],
[0, 0, 1, 1, 1, 1, 0, 0],
[1, 1, 1, 1, 0, 0, 0, 0],
[1, 1, 1, 1, 0, 0, 0, 0],
]))
]) for i in range(3)
], categories=['a', 'b', 'c'])
with self.subTest('upscale'):
actual = transforms.ResizeTransform(small_dataset, width=8, height=8)
compare_datasets(self, big_dataset, actual)
with self.subTest('downscale'):
actual = transforms.ResizeTransform(big_dataset, width=4, height=4)
compare_datasets(self, small_dataset, actual)
def append_polygon_voc(annotations, **kwargs):
annotations.append(
Polygon(
[0, 0, 1, 0, 1, 2, 0, 2],
label=kwargs["label_id"],
id=kwargs["ann_id"] + 1,
attributes=kwargs["attributes"],
group=kwargs["ann_id"],
)) # obj
annotations.append(
Label(kwargs["label_id"], attributes=kwargs["attributes"]))
annotations.append(
Polygon(
[0, 0, 1, 0, 1, 2, 0, 2],
label=kwargs["label_id"] + 3,
group=kwargs["ann_id"],
)) # part
annotations.append(
Label(kwargs["label_id"] + 3, attributes=kwargs["attributes"]))
def append_polygon_labelme(annotations, **kwargs):
annotations.append(
Polygon(
[0, 0, 1, 0, 1, 2, 0, 2],
label=kwargs["label_id"],
id=kwargs["ann_id"],
attributes=kwargs["attributes"],
))
annotations.append(
Label(kwargs["label_id"], attributes=kwargs["attributes"]))
def test_remap_labels(self):
src_dataset = Dataset.from_iterable([
DatasetItem(id=1, annotations=[
# Should be remapped
Label(1),
Bbox(1, 2, 3, 4, label=2),
Mask(image=np.array([1]), label=3),
# Should be deleted
Polygon([1, 1, 2, 2, 3, 4], label=4),
# Should be kept
PolyLine([1, 3, 4, 2, 5, 6]),
Bbox(4, 3, 2, 1, label=5),
])
], categories={
AnnotationType.label: LabelCategories.from_iterable(
f'label{i}' for i in range(6)),
AnnotationType.mask: MaskCategories(
colormap=mask_tools.generate_colormap(6)),
AnnotationType.points: PointsCategories.from_iterable(
[(i, [str(i)]) for i in range(6)])
})
dst_dataset = Dataset.from_iterable([
DatasetItem(id=1, annotations=[
Label(1),
Bbox(1, 2, 3, 4, label=0),
Mask(image=np.array([1]), label=1),
PolyLine([1, 3, 4, 2, 5, 6], label=None),
Bbox(4, 3, 2, 1, label=2),
]),
], categories={
AnnotationType.label: LabelCategories.from_iterable(
['label0', 'label9', 'label5']),
AnnotationType.mask: MaskCategories(colormap={
i: v for i, v in enumerate({
k: v for k, v in mask_tools.generate_colormap(6).items()
if k in { 0, 1, 5 }
}.values())
}),
AnnotationType.points: PointsCategories.from_iterable(
[(0, ['0']), (1, ['1']), (2, ['5'])])
})
actual = transforms.RemapLabels(src_dataset, mapping={
'label1': 'label9', # rename & join with new label9 (from label3)
'label2': 'label0', # rename & join with existing label0
'label3': 'label9', # rename & join with new label9 (from label1)
'label4': '', # delete the label and associated annotations
# 'label5' - unchanged
}, default='keep')
compare_datasets(self, dst_dataset, actual)
def convert_mask(mask):
polygons = mask_tools.mask_to_polygons(mask.image)
return [
Polygon(points=p,
label=mask.label,
z_order=mask.z_order,
id=mask.id,
attributes=mask.attributes,
group=mask.group) for p in polygons
]
def test_mask_to_polygons(self):
source = Dataset.from_iterable([
DatasetItem(id=1, image=np.zeros((5, 10, 3)), annotations=[
Mask(np.array([
[0, 1, 1, 1, 0, 1, 1, 1, 1, 0],
[0, 0, 1, 1, 0, 1, 1, 1, 0, 0],
[0, 0, 0, 1, 0, 1, 1, 0, 0, 0],
[0, 0, 0, 0, 0, 1, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
]),
),
]),
])
expected = Dataset.from_iterable([
DatasetItem(id=1, image=np.zeros((5, 10, 3)), annotations=[
Polygon([1, 0, 3, 2, 3, 0, 1, 0]),
Polygon([5, 0, 5, 3, 8, 0, 5, 0]),
]),
])
actual = transforms.MasksToPolygons(source)
compare_datasets(self, expected, actual)
def test_can_save_dataset_with_cyrillic_and_spaces_in_filename(self):
source_dataset = Dataset.from_iterable([
DatasetItem(id='кириллица с пробелом', subset='train',
image=np.ones((16, 16, 3)),
annotations=[ Polygon([0, 4, 4, 4, 5, 6], label=3) ]
),
], categories=['label_' + str(label) for label in range(10)])
target_dataset = Dataset.from_iterable([
DatasetItem(id='кириллица с пробелом', subset='train',
image=np.ones((16, 16, 3)),
annotations=[
Polygon([0, 4, 4, 4, 5, 6], label=0, id=0,
attributes={ 'occluded': False, 'username': '' }
),
]
),
], categories=['label_3'])
with TestDir() as test_dir:
self._test_save_and_load(
source_dataset,
partial(LabelMeConverter.convert, save_images=True),
test_dir, target_dataset=target_dataset, require_images=True)
def test_can_save_and_load_bboxes_with_no_save_images(self):
expected_dataset = Dataset.from_iterable([
DatasetItem(id=3,
subset='train',
image=np.ones((10, 15, 3)),
annotations=[
Polygon([2, 2, 8, 3, 7, 10, 2, 9],
attributes={'text': 'word_2'}),
Bbox(0, 2, 5, 9, attributes={'text': 'word_3'}),
]),
])
with TestDir() as test_dir:
self._test_save_and_load(
expected_dataset,
partial(IcdarTextLocalizationConverter.convert,
save_images=False), test_dir,
'icdar_text_localization')
def crop_segments(cls, segment_anns, img_width, img_height):
segment_anns = sorted(segment_anns, key=lambda x: x.z_order)
segments = []
for s in segment_anns:
if s.type == AnnotationType.polygon:
segments.append(s.points)
elif s.type == AnnotationType.mask:
if isinstance(s, RleMask):
rle = s.rle
else:
rle = mask_tools.mask_to_rle(s.image)
segments.append(rle)
segments = mask_tools.crop_covered_segments(segments, img_width,
img_height)
new_anns = []
for ann, new_segment in zip(segment_anns, segments):
fields = {
'z_order': ann.z_order,
'label': ann.label,
'id': ann.id,
'group': ann.group,
'attributes': ann.attributes
}
if ann.type == AnnotationType.polygon:
if fields['group'] is None:
fields['group'] = cls._make_group_id(
segment_anns + new_anns, fields['id'])
for polygon in new_segment:
new_anns.append(Polygon(points=polygon, **fields))
else:
rle = mask_tools.mask_to_rle(new_segment)
rle = mask_utils.frPyObjects(rle, *rle['size'])
new_anns.append(RleMask(rle=rle, **fields))
return new_anns
def test_can_import_bboxes(self):
expected_dataset = Dataset.from_iterable([
DatasetItem(id='img_1',
subset='train',
image=np.ones((10, 15, 3)),
annotations=[
Polygon([0, 0, 3, 1, 4, 6, 1, 7],
attributes={'text': 'FOOD'}),
]),
DatasetItem(id='img_2',
subset='train',
image=np.ones((10, 15, 3)),
annotations=[
Bbox(0, 0, 2, 3, attributes={'text': 'RED'}),
Bbox(3, 3, 2, 3, attributes={'text': 'LION'}),
]),
])
dataset = Dataset.import_from(
osp.join(DUMMY_DATASET_DIR, 'text_localization'),
'icdar_text_localization')
compare_datasets(self, expected_dataset, dataset)
def test_can_match_shapes(self):
source0 = Dataset.from_iterable([
DatasetItem(1, annotations=[
# unique
Bbox(1, 2, 3, 4, label=1),
# common
Mask(label=2, z_order=2, image=np.array([
[0, 0, 0, 0],
[0, 0, 0, 0],
[1, 1, 1, 0],
[1, 1, 1, 0],
])),
Polygon([1, 0, 3, 2, 1, 2]),
# an instance with keypoints
Bbox(4, 5, 2, 4, label=2, z_order=1, group=1),
Points([5, 6], label=0, group=1),
Points([6, 8], label=1, group=1),
PolyLine([1, 1, 2, 1, 3, 1]),
]),
], categories=['a', 'b', 'c'])
source1 = Dataset.from_iterable([
DatasetItem(1, annotations=[
# common
Mask(label=2, image=np.array([
[0, 0, 0, 0],
[0, 1, 1, 1],
[0, 1, 1, 1],
[0, 1, 1, 1],
])),
Polygon([0, 2, 2, 0, 2, 1]),
# an instance with keypoints
Bbox(4, 4, 2, 5, label=2, z_order=1, group=2),
Points([5.5, 6.5], label=0, group=2),
Points([6, 8], label=1, group=2),
PolyLine([1, 1.5, 2, 1.5]),
]),
], categories=['a', 'b', 'c'])
source2 = Dataset.from_iterable([
DatasetItem(1, annotations=[
# common
Mask(label=2, z_order=3, image=np.array([
[0, 0, 1, 1],
[0, 1, 1, 1],
[1, 1, 1, 1],
[1, 1, 1, 0],
])),
Polygon([3, 1, 2, 2, 0, 1]),
# an instance with keypoints, one is missing
Bbox(3, 6, 2, 3, label=2, z_order=4, group=3),
Points([4.5, 5.5], label=0, group=3),
PolyLine([1, 1.25, 3, 1, 4, 2]),
]),
], categories=['a', 'b', 'c'])
expected = Dataset.from_iterable([
DatasetItem(1, annotations=[
# unique
Bbox(1, 2, 3, 4, label=1),
# common
# nearest to mean bbox
Mask(label=2, z_order=3, image=np.array([
[0, 0, 0, 0],
[0, 1, 1, 1],
[0, 1, 1, 1],
[0, 1, 1, 1],
])),
Polygon([1, 0, 3, 2, 1, 2]),
# an instance with keypoints
Bbox(4, 5, 2, 4, label=2, z_order=4, group=1),
Points([5, 6], label=0, group=1),
Points([6, 8], label=1, group=1),
PolyLine([1, 1.25, 3, 1, 4, 2]),
]),
], categories=['a', 'b', 'c'])
merger = IntersectMerge(conf={'quorum': 1, 'pairwise_dist': 0.1})
merged = merger([source0, source1, source2])
compare_datasets(self, expected, merged, ignored_attrs={'score'})
self.assertEqual(
[
NoMatchingAnnError(item_id=('1', DEFAULT_SUBSET_NAME),
sources={2}, ann=source0.get('1').annotations[5]),
NoMatchingAnnError(item_id=('1', DEFAULT_SUBSET_NAME),
sources={1, 2}, ann=source0.get('1').annotations[0]),
],
sorted((e for e in merger.errors
if isinstance(e, NoMatchingAnnError)),
key=lambda e: len(e.sources))
)
def _load_annotations(item):
parsed = item['annotations']
loaded = []
for ann in parsed:
ann_id = ann.get('id')
ann_type = AnnotationType[ann['type']]
attributes = ann.get('attributes')
group = ann.get('group')
label_id = ann.get('label_id')
z_order = ann.get('z_order')
points = ann.get('points')
if ann_type == AnnotationType.label:
loaded.append(
Label(label=label_id,
id=ann_id,
attributes=attributes,
group=group))
elif ann_type == AnnotationType.mask:
rle = ann['rle']
rle['counts'] = rle['counts'].encode('ascii')
loaded.append(
RleMask(rle=rle,
label=label_id,
id=ann_id,
attributes=attributes,
group=group,
z_order=z_order))
elif ann_type == AnnotationType.polyline:
loaded.append(
PolyLine(points,
label=label_id,
id=ann_id,
attributes=attributes,
group=group,
z_order=z_order))
elif ann_type == AnnotationType.polygon:
loaded.append(
Polygon(points,
label=label_id,
id=ann_id,
attributes=attributes,
group=group,
z_order=z_order))
elif ann_type == AnnotationType.bbox:
x, y, w, h = ann['bbox']
loaded.append(
Bbox(x,
y,
w,
h,
label=label_id,
id=ann_id,
attributes=attributes,
group=group,
z_order=z_order))
elif ann_type == AnnotationType.points:
loaded.append(
Points(points,
label=label_id,
id=ann_id,
attributes=attributes,
group=group,
z_order=z_order))
elif ann_type == AnnotationType.caption:
caption = ann.get('caption')
loaded.append(
Caption(caption,
id=ann_id,
attributes=attributes,
group=group))
elif ann_type == AnnotationType.cuboid_3d:
loaded.append(
Cuboid3d(ann.get('position'),
ann.get('rotation'),
ann.get('scale'),
label=label_id,
id=ann_id,
attributes=attributes,
group=group))
else:
raise NotImplementedError()
return loaded
def _load_annotations(self, ann, image_info=None, parsed_annotations=None):
if parsed_annotations is None:
parsed_annotations = []
ann_id = ann['id']
attributes = ann.get('attributes', {})
if 'score' in ann:
attributes['score'] = ann['score']
group = ann_id # make sure all tasks' annotations are merged
if self._task is CocoTask.instances or \
self._task is CocoTask.person_keypoints or \
self._task is CocoTask.stuff:
label_id = self._get_label_id(ann)
attributes['is_crowd'] = bool(ann['iscrowd'])
if self._task is CocoTask.person_keypoints:
keypoints = ann['keypoints']
points = []
visibility = []
for x, y, v in take_by(keypoints, 3):
points.append(x)
points.append(y)
visibility.append(v)
parsed_annotations.append(
Points(points,
visibility,
label=label_id,
id=ann_id,
attributes=attributes,
group=group))
segmentation = ann['segmentation']
if segmentation and segmentation != [[]]:
rle = None
if isinstance(segmentation, list):
if not self._merge_instance_polygons:
# polygon - a single object can consist of multiple parts
for polygon_points in segmentation:
parsed_annotations.append(
Polygon(points=polygon_points,
label=label_id,
id=ann_id,
attributes=attributes,
group=group))
else:
# merge all parts into a single mask RLE
rle = self._lazy_merged_mask(segmentation,
image_info['height'],
image_info['width'])
elif isinstance(segmentation['counts'], list):
# uncompressed RLE
img_h = image_info['height']
img_w = image_info['width']
mask_h, mask_w = segmentation['size']
if img_h == mask_h and img_w == mask_w:
rle = self._lazy_merged_mask([segmentation], mask_h,
mask_w)
else:
log.warning(
"item #%s: mask #%s "
"does not match image size: %s vs. %s. "
"Skipping this annotation.", image_info['id'],
ann_id, (mask_h, mask_w), (img_h, img_w))
else:
# compressed RLE
rle = segmentation
if rle:
parsed_annotations.append(
RleMask(rle=rle,
label=label_id,
id=ann_id,
attributes=attributes,
group=group))
else:
x, y, w, h = ann['bbox']
parsed_annotations.append(
Bbox(x,
y,
w,
h,
label=label_id,
id=ann_id,
attributes=attributes,
group=group))
elif self._task is CocoTask.labels:
label_id = self._get_label_id(ann)
parsed_annotations.append(
Label(label=label_id,
id=ann_id,
attributes=attributes,
group=group))
elif self._task is CocoTask.captions:
caption = ann['caption']
parsed_annotations.append(
Caption(caption, id=ann_id, attributes=attributes,
group=group))
else:
raise NotImplementedError()
return parsed_annotations
def test_can_import(self):
img1 = np.ones((77, 102, 3)) * 255
img1[6:32, 7:41] = 0
mask1 = np.zeros((77, 102), dtype=int)
mask1[67:69, 58:63] = 1
mask2 = np.zeros((77, 102), dtype=int)
mask2[13:25, 54:71] = [
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0],
[0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0],
[0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0],
[0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0],
[0, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 0, 0],
[0, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 1, 1, 1, 0, 0, 0],
[0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0],
[0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0],
[0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0],
[0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
]
target_dataset = Dataset.from_iterable([
DatasetItem(id='example_folder/img1', image=img1,
annotations=[
Polygon([43, 34, 45, 34, 45, 37, 43, 37],
label=0, id=0,
attributes={
'occluded': False,
'username': '******'
}
),
Mask(mask1, label=1, id=1,
attributes={
'occluded': False,
'username': '******'
}
),
Polygon([30, 12, 42, 21, 24, 26, 15, 22, 18, 14, 22, 12, 27, 12],
label=2, group=2, id=2,
attributes={
'a1': True,
'occluded': True,
'username': '******'
}
),
Polygon([35, 21, 43, 22, 40, 28, 28, 31, 31, 22, 32, 25],
label=3, group=2, id=3,
attributes={
'kj': True,
'occluded': False,
'username': '******'
}
),
Bbox(13, 19, 10, 11, label=4, group=2, id=4,
attributes={
'hg': True,
'occluded': True,
'username': '******'
}
),
Mask(mask2, label=5, group=1, id=5,
attributes={
'd': True,
'occluded': False,
'username': '******'
}
),
Polygon([64, 21, 74, 24, 72, 32, 62, 34, 60, 27, 62, 22],
label=6, group=1, id=6,
attributes={
'gfd lkj lkj hi': True,
'occluded': False,
'username': '******'
}
),
]
),
], categories=[
'window', 'license plate', 'o1', 'q1', 'b1', 'm1', 'hg',
])
parsed = Dataset.import_from(DUMMY_DATASET_DIR, 'label_me')
compare_datasets(self, expected=target_dataset, actual=parsed)
def test_can_save_and_load(self):
source_dataset = Dataset.from_iterable([
DatasetItem(id='dir1/1', subset='train',
image=np.ones((16, 16, 3)),
annotations=[
Bbox(0, 4, 4, 8, label=2, group=2),
Polygon([0, 4, 4, 4, 5, 6], label=3, attributes={
'occluded': True,
'a1': 'qwe',
'a2': True,
'a3': 123,
'a4': '42', # must be escaped and recognized as string
'escaped': 'a,b. = \\= \\\\ " \\" \\, \\',
}),
Mask(np.array([[0, 1], [1, 0], [1, 1]]), group=2,
attributes={ 'username': '******' }),
Bbox(1, 2, 3, 4, group=3),
Mask(np.array([[0, 0], [0, 0], [1, 1]]), group=3,
attributes={ 'occluded': True }
),
]
),
], categories=['label_' + str(label) for label in range(10)])
target_dataset = Dataset.from_iterable([
DatasetItem(id='dir1/1', subset='train',
image=np.ones((16, 16, 3)),
annotations=[
Bbox(0, 4, 4, 8, label=0, group=2, id=0,
attributes={
'occluded': False, 'username': '',
}
),
Polygon([0, 4, 4, 4, 5, 6], label=1, id=1,
attributes={
'occluded': True, 'username': '',
'a1': 'qwe',
'a2': True,
'a3': 123,
'a4': '42',
'escaped': 'a,b. = \\= \\\\ " \\" \\, \\',
}
),
Mask(np.array([[0, 1], [1, 0], [1, 1]]), group=2,
id=2, attributes={
'occluded': False, 'username': '******'
}
),
Bbox(1, 2, 3, 4, group=1, id=3, attributes={
'occluded': False, 'username': '',
}),
Mask(np.array([[0, 0], [0, 0], [1, 1]]), group=1,
id=4, attributes={
'occluded': True, 'username': ''
}
),
]
),
], categories=['label_2', 'label_3'])
with TestDir() as test_dir:
self._test_save_and_load(
source_dataset,
partial(LabelMeConverter.convert, save_images=True),
test_dir, target_dataset=target_dataset, require_images=True)
def test_can_save_and_load(self):
src_label_cat = LabelCategories(attributes={'occluded', 'common'})
for i in range(10):
src_label_cat.add(str(i))
src_label_cat.items[2].attributes.update(['a1', 'a2', 'empty'])
source_dataset = Dataset.from_iterable([
DatasetItem(id=0, subset='s1', image=np.zeros((5, 10, 3)),
annotations=[
Polygon([0, 0, 4, 0, 4, 4],
label=1, group=4,
attributes={ 'occluded': True, 'common': 't' }),
Points([1, 1, 3, 2, 2, 3],
label=2,
attributes={ 'a1': 'x', 'a2': 42, 'empty': '',
'unknown': 'bar' }),
Label(1),
Label(2, attributes={ 'a1': 'y', 'a2': 44 }),
]
),
DatasetItem(id=1, subset='s1',
annotations=[
PolyLine([0, 0, 4, 0, 4, 4],
label=3, id=4, group=4),
Bbox(5, 0, 1, 9,
label=3, id=4, group=4),
]
),
DatasetItem(id=2, subset='s2', image=np.ones((5, 10, 3)),
annotations=[
Polygon([0, 0, 4, 0, 4, 4], z_order=1,
label=3, group=4,
attributes={ 'occluded': False }),
PolyLine([5, 0, 9, 0, 5, 5]), # will be skipped as no label
]
),
DatasetItem(id=3, subset='s3', image=Image(
path='3.jpg', size=(2, 4))),
], categories={ AnnotationType.label: src_label_cat })
target_label_cat = LabelCategories(
attributes={'occluded'}) # unable to represent a common attribute
for i in range(10):
target_label_cat.add(str(i), attributes={'common'})
target_label_cat.items[2].attributes.update(['a1', 'a2', 'empty', 'common'])
target_dataset = Dataset.from_iterable([
DatasetItem(id=0, subset='s1', image=np.zeros((5, 10, 3)),
annotations=[
Polygon([0, 0, 4, 0, 4, 4],
label=1, group=4,
attributes={ 'occluded': True, 'common': 't' }),
Points([1, 1, 3, 2, 2, 3],
label=2,
attributes={ 'occluded': False, 'empty': '',
'a1': 'x', 'a2': '42' }),
Label(1),
Label(2, attributes={ 'a1': 'y', 'a2': '44' }),
], attributes={'frame': 0}
),
DatasetItem(id=1, subset='s1',
annotations=[
PolyLine([0, 0, 4, 0, 4, 4],
label=3, group=4,
attributes={ 'occluded': False }),
Bbox(5, 0, 1, 9,
label=3, group=4,
attributes={ 'occluded': False }),
], attributes={'frame': 1}
),
DatasetItem(id=2, subset='s2', image=np.ones((5, 10, 3)),
annotations=[
Polygon([0, 0, 4, 0, 4, 4], z_order=1,
label=3, group=4,
attributes={ 'occluded': False }),
], attributes={'frame': 0}
),
DatasetItem(id=3, subset='s3', image=Image(
path='3.jpg', size=(2, 4)),
attributes={'frame': 0}),
], categories={ AnnotationType.label: target_label_cat })
with TestDir() as test_dir:
self._test_save_and_load(source_dataset,
partial(CvatConverter.convert, save_images=True), test_dir,
target_dataset=target_dataset)
def _load_localization_items(self):
items = {}
image_dir = osp.join(self._path, IcdarPath.IMAGES_DIR)
if osp.isdir(image_dir):
images = {
osp.splitext(osp.relpath(p, image_dir))[0].replace('\\', '/'): p
for p in find_images(image_dir, recursive=True)
}
else:
images = {}
for path in glob.iglob(
osp.join(self._path, '**', '*.txt'), recursive=True):
item_id = osp.splitext(osp.relpath(path, self._path))[0]
if osp.basename(item_id).startswith('gt_'):
item_id = osp.join(osp.dirname(item_id), osp.basename(item_id)[3:])
item_id = item_id.replace('\\', '/')
if item_id not in items:
items[item_id] = DatasetItem(item_id, subset=self._subset,
image=images.get(item_id))
annotations = items[item_id].annotations
with open(path, encoding='utf-8') as f:
for line in f:
line = line.strip()
objects = line.split('\"')
if 1 < len(objects):
if len(objects) == 3:
text = objects[1]
else:
raise Exception("Line %s: unexpected number "
"of quotes in filename" % line)
else:
text = ''
objects = objects[0].split()
if len(objects) == 1:
objects = objects[0].split(',')
if 8 <= len(objects):
points = [float(p) for p in objects[:8]]
attributes = {}
if 0 < len(text):
attributes['text'] = text
elif len(objects) == 9:
text = objects[8]
attributes['text'] = text
annotations.append(
Polygon(points, attributes=attributes))
elif 4 <= len(objects):
x = float(objects[0])
y = float(objects[1])
w = float(objects[2]) - x
h = float(objects[3]) - y
attributes = {}
if 0 < len(text):
attributes['text'] = text
elif len(objects) == 5:
text = objects[4]
attributes['text'] = text
annotations.append(
Bbox(x, y, w, h, attributes=attributes))
return items
def _load_items(self, subset):
labels = self._categories.setdefault(AnnotationType.label,
LabelCategories())
path = osp.join(self._path, subset)
images = [i for i in find_images(path, recursive=True)]
for image_path in sorted(images):
item_id = osp.splitext(osp.relpath(image_path, path))[0]
if Ade20k2020Path.MASK_PATTERN.fullmatch(osp.basename(item_id)):
continue
item_annotations = []
item_info = self._load_item_info(image_path)
for item in item_info:
label_idx = labels.find(item['label_name'])[0]
if label_idx is None:
labels.add(item['label_name'])
mask_path = osp.splitext(image_path)[0] + '_seg.png'
max_part_level = max([p['part_level'] for p in item_info])
for part_level in range(max_part_level + 1):
if not osp.exists(mask_path):
log.warning('Can`t find part level %s mask for %s' \
% (part_level, image_path))
continue
mask = lazy_image(mask_path, loader=self._load_class_mask)
mask = CompiledMask(instance_mask=mask)
classes = {(v['class_idx'], v['label_name'])
for v in item_info if v['part_level'] == part_level}
for class_idx, label_name in classes:
label_id = labels.find(label_name)[0]
item_annotations.append(
Mask(label=label_id,
id=class_idx,
image=mask.lazy_extract(class_idx),
group=class_idx,
z_order=part_level))
mask_path = osp.splitext(image_path)[0] \
+ '_parts_%s.png' % (part_level + 1)
for item in item_info:
instance_path = osp.join(osp.dirname(image_path),
item['instance_mask'])
if not osp.isfile(instance_path):
log.warning('Can`t find instance mask: %s' % instance_path)
continue
mask = lazy_image(instance_path,
loader=self._load_instance_mask)
mask = CompiledMask(instance_mask=mask)
label_id = labels.find(item['label_name'])[0]
instance_id = item['id']
attributes = {k: True for k in item['attributes']}
polygon_points = item['polygon_points']
item_annotations.append(
Mask(label=label_id,
image=mask.lazy_extract(1),
id=instance_id,
attributes=attributes,
z_order=item['part_level'],
group=instance_id))
if (len(item['polygon_points']) % 2 == 0 \
and 3 <= len(item['polygon_points']) // 2):
item_annotations.append(
Polygon(polygon_points,
label=label_id,
attributes=attributes,
id=instance_id,
z_order=item['part_level'],
group=instance_id))
self._items.append(
DatasetItem(item_id,
subset=subset,
image=image_path,
annotations=item_annotations))