def test_semantic_segmentation_labelling_sparse():
"""Test `semantic_segmentation_labelling` function in `generator` module by considering a
sparse labelling, *i.e.* the array contains unknown values (to mimic the non-evaluated label
situations)
* as a preliminary verification, check if passing string labels raises an AttributeError
exception
* test if output shape is input shape + an additional dimension given by the
`label_ids` length
* test if both representation provides the same information (native array on the
first hand and its one-hot version on the second hand)
"""
a = np.array([[[1, 1, 3, 1], [3, 3, 1, 1], [3, 4, 3, 1]],
[[1, 1, 0, 0], [3, 4, 0, 1], [1, 1, 0, 0]]])
labels = [0, 2, 3]
asum, _ = np.histogram(a.reshape(-1), range=(np.amin(a), np.amax(a)))
with pytest.raises(ValueError):
b = generator.semantic_segmentation_labelling(a, ['0', '2', '3'])
b = generator.semantic_segmentation_labelling(a, labels)
assert len(labels) != np.amax(a) - np.amin(a) + 1
assert b.shape == (a.shape[0], a.shape[1], a.shape[2], len(labels))
assert b.tolist() == [[[[False, False, False], [False, False, False],
[False, False, True], [False, False, False]],
[[False, False, True], [False, False, True],
[False, False, False], [False, False, False]],
[[False, False, True], [False, False, False],
[False, False, True], [False, False, False]]],
[[[False, False, False], [False, False, False],
[True, False, False], [True, False, False]],
[[False, False, True], [False, False, False],
[True, False, False], [False, False, False]],
[[False, False, False], [False, False, False],
[True, False, False], [True, False, False]]]]
def test_semantic_segmentation_labelling_sparse():
"""Test `semantic_segmentation_labelling` function in `generator` module by
considering a sparse labelling, *i.e.* the array contains unknown values
(to mimic the non-evaluated label situations)
* as a preliminary verification, check if passing string labels raises an
AttributeError exception
* test if output shape is input shape + an additional dimension given by
the `label_ids` length
* test if both representation provides the same information (native array
on the first hand and its one-hot version on the second hand)
"""
a = np.array([
[
[[200, 10, 10], [200, 10, 10], [200, 200, 200]],
[[200, 200, 200], [200, 200, 200], [200, 10, 10]],
[[200, 200, 200], [100, 100, 100], [200, 200, 200]],
],
[
[[200, 10, 10], [200, 10, 10], [10, 10, 200]],
[[200, 200, 200], [100, 100, 100], [10, 10, 200]],
[[200, 10, 10], [200, 10, 10], [10, 10, 200]],
],
])
asum, _ = np.histogram(a.reshape(-1), range=(np.amin(a), np.amax(a)))
wrong_config = [
{
"id": "0",
"color": [10, 10, 200],
"is_evaluate": True
},
{
"id": "2",
"color": [10, 200, 10],
"is_evaluate": True
},
{
"id": "3",
"color": [200, 200, 200],
"is_evaluate": True
},
]
with pytest.raises(ValueError):
b = generator.semantic_segmentation_labelling(a, wrong_config)
config = [
{
"id": 0,
"color": [10, 10, 200],
"is_evaluate": True
},
{
"id": 2,
"color": [10, 200, 10],
"is_evaluate": True
},
{
"id": 3,
"color": [200, 200, 200],
"is_evaluate": True
},
]
labels = [item["id"] for item in config]
b = generator.semantic_segmentation_labelling(a, config)
assert len(labels) != np.amax(a) - np.amin(a) + 1
assert b.shape == (a.shape[0], a.shape[1], a.shape[2], len(labels))
assert b.tolist() == [
[
[
[False, False, False],
[False, False, False],
[False, False, True],
],
[
[False, False, True],
[False, False, True],
[False, False, False],
],
[
[False, False, True],
[False, False, False],
[False, False, True],
],
],
[
[
[False, False, False],
[False, False, False],
[True, False, False],
],
[
[False, False, True],
[False, False, False],
[True, False, False],
],
[
[False, False, False],
[False, False, False],
[True, False, False],
],
],
]
def test_semantic_segmentation_labelling_concise():
"""Test `semantic_segmentation_labelling` function in `generator` module by considering a
concise labelling, *i.e.* the labels correspond to array values
* as a preliminary verification, check if passing string labels raises an AttributeError
exception
* test if output shape is input shape + an additional dimension given by the
`label_ids` length
* test if both representation provides the same information (native array on the
first hand and its one-hot version on the second hand)
"""
a = np.array([[[[200, 10, 10], [200, 10, 10], [200, 200, 200]],
[[200, 200, 200], [200, 200, 200], [200, 10, 10]],
[[200, 200, 200], [200, 200, 200], [200, 200, 200]]],
[[[200, 10, 10], [200, 10, 10], [10, 10, 200]],
[[10, 200, 10], [10, 200, 10], [10, 10, 200]],
[[200, 10, 10], [200, 10, 10], [10, 10, 200]]]])
labels = np.unique(a.reshape(-1, 3), axis=0).tolist()
wrong_config = [{
'id': '0',
'color': [10, 10, 200],
'is_evaluate': True
}, {
'id': '1',
'color': [200, 10, 10],
'is_evaluate': True
}, {
'id': '2',
'color': [10, 200, 10],
'is_evaluate': True
}, {
'id': '3',
'color': [200, 200, 200],
'is_evaluate': True
}]
asum, _ = np.histogram(a.reshape(-1), range=(np.amin(a), np.amax(a)))
with pytest.raises(ValueError):
b = generator.semantic_segmentation_labelling(a, wrong_config)
config = [{
'id': 0,
'color': [10, 10, 200],
'is_evaluate': True
}, {
'id': 1,
'color': [200, 10, 10],
'is_evaluate': True
}, {
'id': 2,
'color': [10, 200, 10],
'is_evaluate': True
}, {
'id': 3,
'color': [200, 200, 200],
'is_evaluate': True
}]
b = generator.semantic_segmentation_labelling(a, config)
assert b.shape == (a.shape[0], a.shape[1], a.shape[2], len(labels))
assert b.tolist() == [[[[False, True, False, False],
[False, True, False, False],
[False, False, False, True]],
[[False, False, False, True],
[False, False, False, True],
[False, True, False, False]],
[[False, False, False, True],
[False, False, False, True],
[False, False, False, True]]],
[[[False, True, False, False],
[False, True, False, False],
[True, False, False, False]],
[[False, False, True, False],
[False, False, True, False],
[True, False, False, False]],
[[False, True, False, False],
[False, True, False, False],
[True, False, False, False]]]]