def test_dataset_with_guessed_labelmap(self):
class SrcExtractor(TestExtractorBase):
def __iter__(self):
yield DatasetItem(id=1,
annotations=[
Bbox(2, 3, 4, 5, label=0, id=1),
Bbox(1, 2, 3, 4, label=1, id=2),
])
def categories(self):
label_cat = LabelCategories()
label_cat.add(VOC.VocLabel(1).name)
label_cat.add('non_voc_label')
return {
AnnotationType.label: label_cat,
}
class DstExtractor(TestExtractorBase):
def __iter__(self):
yield DatasetItem(id=1,
annotations=[
Bbox(2,
3,
4,
5,
label=self._label(
VOC.VocLabel(1).name),
id=1,
group=1,
attributes={
'truncated': False,
'difficult': False,
'occluded': False,
}),
Bbox(1,
2,
3,
4,
label=self._label('non_voc_label'),
id=2,
group=2,
attributes={
'truncated': False,
'difficult': False,
'occluded': False,
}),
])
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)
with TestDir() as test_dir:
self._test_save_and_load(SrcExtractor(),
VocConverter(label_map='guess'),
test_dir,
target_dataset=DstExtractor())
def test_dataset_with_source_labelmap_undefined(self):
class SrcExtractor(TestExtractorBase):
def __iter__(self):
yield DatasetItem(id=1,
annotations=[
Bbox(2, 3, 4, 5, label=0, id=1),
Bbox(1, 2, 3, 4, label=1, id=2),
])
def categories(self):
label_cat = LabelCategories()
label_cat.add('Label_1')
label_cat.add('label_2')
return {
AnnotationType.label: label_cat,
}
class DstExtractor(TestExtractorBase):
def __iter__(self):
yield DatasetItem(id=1,
annotations=[
Bbox(2,
3,
4,
5,
label=self._label('Label_1'),
id=1,
group=1,
attributes={
'truncated': False,
'difficult': False,
'occluded': False,
}),
Bbox(1,
2,
3,
4,
label=self._label('label_2'),
id=2,
group=2,
attributes={
'truncated': False,
'difficult': False,
'occluded': False,
}),
])
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)
with TestDir() as test_dir:
self._test_save_and_load(SrcExtractor(),
VocConverter(label_map='source'),
test_dir,
target_dataset=DstExtractor())
def test_background_masks_dont_introduce_instances_but_cover_others(self):
dataset = Dataset.from_iterable([
DatasetItem(1, image=np.zeros((4, 1, 1)), annotations=[
Mask([1, 1, 1, 1], label=1, attributes={'z_order': 1}),
Mask([0, 0, 1, 1], label=2, attributes={'z_order': 2}),
Mask([0, 0, 1, 1], label=0, attributes={'z_order': 3}),
])
], categories=['background', 'a', 'b'])
with TestDir() as test_dir:
VocConverter.convert(dataset, test_dir, apply_colormap=False)
cls_mask = load_mask(
osp.join(test_dir, 'SegmentationClass', '1.png'))
inst_mask = load_mask(
osp.join(test_dir, 'SegmentationObject', '1.png'))
self.assertTrue(np.array_equal([0, 1], np.unique(cls_mask)))
self.assertTrue(np.array_equal([0, 1], np.unique(inst_mask)))
def test_can_save_dataset_with_image_info(self):
class TestExtractor(TestExtractorBase):
def __iter__(self):
return iter([
DatasetItem(id=1, image=Image(path='1.jpg', size=(10, 15))),
])
with TestDir() as test_dir:
self._test_save_and_load(TestExtractor(),
VocConverter(label_map='voc'), test_dir)
def test_dataset_with_voc_labelmap(self):
class SrcExtractor(TestExtractorBase):
def __iter__(self):
yield DatasetItem(id=1,
annotations=[
Bbox(2,
3,
4,
5,
label=self._label('cat'),
id=1),
Bbox(1,
2,
3,
4,
label=self._label('non_voc_label'),
id=2),
])
def categories(self):
label_cat = LabelCategories()
label_cat.add(VOC.VocLabel.cat.name)
label_cat.add('non_voc_label')
return {
AnnotationType.label: label_cat,
}
class DstExtractor(TestExtractorBase):
def __iter__(self):
yield DatasetItem(
id=1,
annotations=[
# drop non voc label
Bbox(2,
3,
4,
5,
label=self._label('cat'),
id=1,
group=1,
attributes={
'truncated': False,
'difficult': False,
'occluded': False,
}),
])
def categories(self):
return VOC.make_voc_categories()
with TestDir() as test_dir:
self._test_save_and_load(SrcExtractor(),
VocConverter(label_map='voc'),
test_dir,
target_dataset=DstExtractor())
def test_relative_paths(self):
class TestExtractor(TestExtractorBase):
def __iter__(self):
return iter([
DatasetItem(id='1', image=np.ones((4, 2, 3))),
DatasetItem(id='subdir1/1', image=np.ones((2, 6, 3))),
DatasetItem(id='subdir2/1', image=np.ones((5, 4, 3))),
])
with TestDir() as test_dir:
self._test_save_and_load(TestExtractor(),
VocConverter(label_map='voc', save_images=True), test_dir)
def test_can_save_dataset_with_images(self):
class TestExtractor(TestExtractorBase):
def __iter__(self):
return iter([
DatasetItem(id=1, subset='a', image=np.ones([4, 5, 3])),
DatasetItem(id=2, subset='a', image=np.ones([5, 4, 3])),
DatasetItem(id=3, subset='b', image=np.ones([2, 6, 3])),
])
with TestDir() as test_dir:
self._test_save_and_load(TestExtractor(),
VocConverter(label_map='voc', save_images=True), test_dir)
def test_dataset_with_fixed_labelmap(self):
class SrcExtractor(TestExtractorBase):
def __iter__(self):
yield DatasetItem(id=1, annotations=[
Bbox(2, 3, 4, 5, label=0, id=1),
Bbox(1, 2, 3, 4, label=1, id=2, group=2,
attributes={'act1': True}),
Bbox(2, 3, 4, 5, label=2, id=3, group=2),
Bbox(2, 3, 4, 6, label=3, id=4, group=2),
])
def categories(self):
label_cat = LabelCategories()
label_cat.add('foreign_label')
label_cat.add('label', attributes=['act1', 'act2'])
label_cat.add('label_part1')
label_cat.add('label_part2')
return {
AnnotationType.label: label_cat,
}
label_map = {
'label': [None, ['label_part1', 'label_part2'], ['act1', 'act2']]
}
class DstExtractor(TestExtractorBase):
def __iter__(self):
yield DatasetItem(id=1, annotations=[
Bbox(1, 2, 3, 4, label=0, id=1, group=1,
attributes={
'act1': True,
'act2': False,
'truncated': False,
'difficult': False,
'occluded': False,
}
),
Bbox(2, 3, 4, 5, label=1, group=1),
Bbox(2, 3, 4, 6, label=2, group=1),
])
def categories(self):
return VOC.make_voc_categories(label_map)
with TestDir() as test_dir:
self._test_save_and_load(
SrcExtractor(), VocConverter(label_map=label_map),
test_dir, target_dataset=DstExtractor())
def test_can_save_dataset_with_no_subsets(self):
class TestExtractor(TestExtractorBase):
def __iter__(self):
return iter([
DatasetItem(id=1, annotations=[
Label(2),
Label(3),
]),
DatasetItem(id=2, annotations=[
Label(3),
]),
])
with TestDir() as test_dir:
self._test_save_and_load(TestExtractor(),
VocConverter(label_map='voc'), test_dir)
