def test_resize_img_to_arbitrary_size(img, mask, resize_to):
# Setting up the data
kpts_data = np.array([[0, 0], [0, 2], [2, 2], [2, 0]]).reshape(
(4, 2)) # Top left corner
kpts = slc.Keypoints(kpts_data.copy(), frame=(img.shape[:2]))
dc = slc.DataContainer((
img,
mask,
kpts,
), "IMP")
transf = slt.Resize(resize_to=resize_to)
res = transf(dc).data
if isinstance(resize_to, int):
resize_to = (resize_to, resize_to)
scales = tuple(resize_to[i] / img.shape[i] for i in range(img.ndim - 1))
assert transf.resize_to == resize_to
np.testing.assert_array_equal(res[0].shape[:-1], resize_to)
np.testing.assert_array_equal(res[1].shape, resize_to)
np.testing.assert_array_equal(res[2].frame, resize_to)
kpts_data = kpts_data.astype(float)
kpts_data = (kpts_data * np.asarray(scales)[None, ])
kpts_data = kpts_data.astype(int)
assert np.array_equal(res[2].data, kpts_data)
def test_reflective_padding_cant_be_applied_to_kpts():
kpts_data = np.array([[0, 0], [0, 1], [1, 0], [2, 0]]).reshape((4, 2))
kpts = slc.Keypoints(kpts_data, 3, 4)
dc = slc.DataContainer((1, kpts), "LP")
trf = slt.Pad(pad_to=(10, 10), padding="r")
with pytest.raises(ValueError):
trf(dc)
def test_rotate_90_img_mask_keypoints_destructive(img_3x3, mask_3x3, transform_settings, ignore_state):
# Setting up the data
kpts_data = np.array([[0, 0], [0, 2], [2, 2], [2, 0]]).reshape((4, 2))
kpts = slc.Keypoints(kpts_data, 3, 3)
img, mask = img_3x3, mask_3x3
H, W = mask.shape
dc = slc.DataContainer((img, mask, kpts, 1), "IMPL", transform_settings=copy.deepcopy(transform_settings),)
# Defining the 90 degrees transform (clockwise)
stream = slt.Rotate(angle_range=(90, 90), p=1, ignore_state=ignore_state)
dc_res = stream(dc)
img_res, _, _ = dc_res[0]
mask_res, _, _ = dc_res[1]
kpts_res, _, _ = dc_res[2]
label_res, _, _ = dc_res[3]
M = cv2.getRotationMatrix2D((W // 2, H // 2), -90, 1)
img_inter = ALLOWED_INTERPOLATIONS["bicubic"]
img_pad = ALLOWED_PADDINGS["z"]
if transform_settings is not None:
img_inter = ALLOWED_INTERPOLATIONS[transform_settings[0]["interpolation"]]
img_pad = ALLOWED_PADDINGS[transform_settings[0]["padding"]]
expected_img_res = cv2.warpAffine(img, M, (W, H), flags=img_inter, borderMode=img_pad).reshape((H, W, 1))
expected_mask_res = cv2.warpAffine(mask, M, (W, H))
expected_kpts_res = np.array([[2, 0], [0, 0], [0, 2], [2, 2]]).reshape((4, 2))
assert np.array_equal(expected_img_res, img_res)
assert np.array_equal(expected_mask_res, mask_res)
np.testing.assert_array_almost_equal(expected_kpts_res, kpts_res.data)
assert label_res == 1
def test_transform_returns_original_data_if_not_in_specified_indices(
trf, img_3x3_rgb):
img_3x3 = img_3x3_rgb * 128
kpts_data = np.array([[0, 0], [0, 2], [2, 2], [2, 0]]).reshape((4, 2))
kpts = slc.Keypoints(kpts_data.copy(), 3, 3)
dc = slc.DataContainer((img_3x3.copy(), img_3x3.copy(), img_3x3.copy(),
img_3x3.copy(), 1, kpts, 2), "IIIILPL")
kwargs = {"p": 1, "data_indices": (0, 1, 4)}
if class_accepts(trf, "gain_range"):
kwargs["gain_range"] = (0.7, 0.9)
if class_accepts(trf, "brightness_range"):
kwargs["brightness_range"] = (10, 20)
if class_accepts(trf, "h_range"):
kwargs["h_range"] = (50, 50)
kwargs["s_range"] = (50, 50)
if class_accepts(trf, "h_range"):
kwargs["h_range"] = (50, 50)
kwargs["s_range"] = (50, 50)
trf = trf(**kwargs)
res = trf(dc)
assert np.linalg.norm(res.data[0] - img_3x3) > 0
assert np.linalg.norm(res.data[1] - img_3x3) > 0
np.testing.assert_array_equal(res.data[2], img_3x3)
np.testing.assert_array_equal(res.data[3], img_3x3)
assert res.data[-1] == 2
np.testing.assert_array_equal(res.data[5].data, kpts_data)
def test_resize_img_to_arbitrary_size(img, mask, resize_to):
# Setting up the data
kpts_data = np.array([[0, 0], [0, 2], [2, 2], [2, 0]]).reshape(
(4, 2)) # Top left corner
kpts = slc.Keypoints(kpts_data.copy(), img.shape[0], img.shape[1])
dc = slc.DataContainer((
img,
mask,
kpts,
), "IMP")
transf = slt.Resize(resize_to=resize_to)
res = transf(dc).data
if isinstance(resize_to, int):
resize_to = (resize_to, resize_to)
scale_x = resize_to[0] / img.shape[1]
scale_y = resize_to[1] / img.shape[0]
assert transf.resize_to == resize_to
assert (res[0].shape[0] == resize_to[1]) and (res[0].shape[1]
== resize_to[0])
assert (res[1].shape[0] == resize_to[1]) and (res[1].shape[1]
== resize_to[0])
assert (res[2].height == resize_to[1]) and (res[2].width == resize_to[0])
kpts_data = kpts_data.astype(float)
kpts_data[:, 0] *= scale_x
kpts_data[:, 1] *= scale_y
kpts_data = kpts_data.astype(int)
assert np.array_equal(res[2].data, kpts_data)
def test_2x2_pad_to_20x20_center_crop_2x2(pad_size, crop_size, img_2x2,
mask_2x2):
# Setting up the data
kpts_data = np.array([[0, 0], [0, 1], [1, 1], [1, 0]]).reshape((4, 2))
kpts = slc.Keypoints(kpts_data, 2, 2)
img, mask = img_2x2, mask_2x2
dc = slc.DataContainer((
img,
mask,
kpts,
), "IMP")
stream = slc.Stream(
[slt.Pad(pad_to=pad_size),
slt.Crop(crop_to=crop_size)])
res = stream(dc, return_torch=False)
assert (res[0][0].shape[0] == 2) and (res[0][0].shape[1] == 2)
assert (res[1][0].shape[0] == 2) and (res[1][0].shape[1] == 2)
assert (res[2][0].height == 2) and (res[2][0].width == 2)
assert np.array_equal(res[0][0], img)
assert np.array_equal(res[1][0], mask)
assert np.array_equal(res[2][0].data, kpts_data)
def test_interpolation_or_padding_settings_for_labels_or_keypoints(setting):
kpts = slc.Keypoints(pts=np.array([[0, 0], [0, 2], [2, 2],
[2, 0]]).reshape((4, 2)),
frame=(3, 3))
with pytest.raises(TypeError):
slc.DataContainer(data=(kpts, ),
fmt="P",
transform_settings={0: setting})
def test_create_4_keypoints_change_frame():
kpts_data = np.array([[0, 0], [0, 1], [1, 0], [1, 1]]).reshape((4, 2))
kpts = slc.Keypoints(kpts_data, frame=(3, 4))
kpts.frame = (2, 2)
assert kpts.frame[0] == 2
assert kpts.frame[1] == 2
assert np.array_equal(kpts_data, kpts.data)
def test_create_4_keypoints_change_frame():
kpts_data = np.array([[0, 0], [0, 1], [1, 0], [1, 1]]).reshape((4, 2))
kpts = slc.Keypoints(kpts_data, 3, 4)
kpts.height = 2
kpts.width = 2
assert kpts.height == 2
assert kpts.width == 2
assert np.array_equal(kpts_data, kpts.data)
def test_keypoints_vertical_flip_within_stream():
kpts_data = np.array([[0, 0], [0, 1], [1, 0], [1, 1]]).reshape((4, 2))
kpts = slc.Keypoints(kpts_data, 2, 2)
stream = slc.Stream([slt.Flip(p=1, axis=0)])
dc = slc.DataContainer((kpts,), "P")
dc_res = stream(dc, return_torch=False)
assert np.array_equal(dc_res[0][0].data, np.array([[0, 1], [0, 0], [1, 1], [1, 0]]).reshape((4, 2)))
def test_keypoint_jitter_works_correctly(jitter_x, jitter_y, exp_x, exp_y):
kpts_data = np.array([[1, 1],]).reshape((1, 2))
kpts = slc.Keypoints(kpts_data.copy(), 2, 2)
dc = slc.DataContainer((kpts,), "P")
trf = slc.Stream([slt.KeypointsJitter(p=1, dx_range=(jitter_x, jitter_x), dy_range=(jitter_y, jitter_y))])
dc_res = trf(dc, return_torch=False)
assert np.array_equal(dc_res.data[0].data, np.array([exp_x, exp_y]).reshape((1, 2)))
def test_pad_crop_resize_dont_change_data_when_parameters_are_not_set(img_3x3, mask_3x3, trf):
# Setting up the data
kpts_data = np.array([[0, 0], [0, 2], [2, 2], [2, 0]]).reshape((4, 2))
kpts = slc.Keypoints(kpts_data, 3, 3)
img, mask = img_3x3, mask_3x3
dc = slc.DataContainer((img, mask, kpts,), "IMP")
res = trf()(dc, return_torch=False)
assert dc == res
def test_gaussian_noise_no_image_throws_value_error():
trf = slt.Noise(p=1)
# Setting up the data
kpts_data = np.array([[0, 0], [0, 5], [1, 3], [2, 0]]).reshape((4, 2))
kpts = slc.Keypoints(kpts_data, 6, 6)
dc = slc.DataContainer((kpts, ), "P")
with pytest.raises(ValueError):
trf(dc)
def test_data_container_keypoints_rescale_to_torch():
kpts_data = np.array([[100, 20], [1023, 80], [20, 20],
[100, 700]]).reshape((4, 2))
kpts = slc.Keypoints(kpts_data, 768, 1024)
ppl = slc.Stream()
k, label = ppl({'keypoints': kpts, 'label': 1}, as_dict=False)
assert isinstance(k, torch.FloatTensor)
np.testing.assert_almost_equal(k.max().item() * 1023, 1023)
np.testing.assert_almost_equal(k.min().item() * 1023, 20)
assert label == 1
def test_keypoints_get_set():
kpts_data = np.array([[0, 0], [0, 1], [1, 0], [1, 1]]).reshape((4, 2))
kpts = slc.Keypoints(kpts_data, 3, 4)
assert np.array_equal(kpts[0], np.array([0, 0]))
kpts[0] = np.array([2, 2])
assert np.array_equal(kpts[0], np.array([2, 2]))
with pytest.raises(TypeError):
kpts[0] = [2, 2]
def test_image_trfs_dont_change_mask_labels_kpts(trf_cls, trf_params, img_3x4, mask_3x4):
trf = trf_cls(**trf_params)
kpts_data = np.array([[0, 0], [0, 1], [1, 0], [2, 0]]).reshape((4, 2))
kpts = slc.Keypoints(kpts_data, 3, 4)
dc = slc.DataContainer((img_3x4, mask_3x4, kpts, 1), "IMPL")
dc_res = trf(dc)
assert np.all(dc.data[1] == dc_res.data[1])
assert np.all(dc.data[2].data == dc_res.data[2].data)
assert dc.data[3] == dc_res.data[3]
def test_keypoints_vertical_flip():
kpts_data = np.array([[0, 0], [0, 1], [1, 0], [1, 1]]).reshape((4, 2))
kpts = slc.Keypoints(kpts_data, frame=(2, 2))
stream = slt.Flip(p=1, axis=0)
dc = slc.DataContainer((kpts, ), "P")
dc_res = stream(dc)
assert np.array_equal(
dc_res[0][0].data,
np.array([[0, 1], [0, 0], [1, 1], [1, 0]]).reshape((4, 2)))
def test_keypoints_assert_reflective(img_3x3, mask_3x3):
# Setting up the data
kpts_data = np.array([[0, 0], [0, 2], [2, 2], [2, 0]]).reshape((4, 2))
kpts = slc.Keypoints(kpts_data, 3, 3)
img, mask = img_3x3, mask_3x3
dc = slc.DataContainer((img, mask, kpts,), "IMP")
# Defining the 90 degrees transform (clockwise)
stream = slt.Rotate(angle_range=(20, 20), p=1, padding="r")
with pytest.raises(ValueError):
stream(dc)
def test_create_4_keypoints_change_grid_and_frame():
kpts_data = np.array([[0, 0], [0, 1], [1, 0], [1, 1]]).reshape((4, 2))
kpts = slc.Keypoints(kpts_data, frame=(3, 4))
kpts_data_new = np.array([[0, 0], [0, 1], [1, 0], [1, 1],
[0.5, 0.5]]).reshape((5, 2))
kpts.frame = (2, 2)
kpts.pts = kpts_data_new
assert kpts.frame[0] == 2
assert kpts.frame[1] == 2
assert np.array_equal(kpts_data_new, kpts.pts)
def test_create_4_keypoints_change_grid_and_frame():
kpts_data = np.array([[0, 0], [0, 1], [1, 0], [1, 1]]).reshape((4, 2))
kpts = slc.Keypoints(kpts_data, 3, 4)
kpts_data_new = np.array([[0, 0], [0, 1], [1, 0], [1, 1],
[0.5, 0.5]]).reshape((5, 2))
kpts.height = 2
kpts.width = 2
kpts.pts = kpts_data_new
assert kpts.height == 2
assert kpts.width == 2
assert np.array_equal(kpts_data_new, kpts.pts)
def test_data_container_from_and_to_dict(img_3x4, mask_3x4, order, presence):
img, mask = img_3x4, mask_3x4
kpts_data = np.array([[0, 0], [0, 1], [1, 0], [1, 1]]).reshape(
(4, 2)).astype(float)
dc, dc_reordered = generate_data_container_based_on_presence(
img, mask, kpts_data, order, presence)
assert dc == dc_reordered
# Now we will also test whether conversion to dict and back works well.
tensor_dict = dc_reordered.to_torch(as_dict=True,
normalize=False,
scale_keypoints=False)
for k in tensor_dict:
if isinstance(tensor_dict[k], (list, tuple)):
tmp = []
for el in tensor_dict[k]:
tmp.append(el.numpy() if isinstance(el, torch.Tensor) else el)
if 'imag' in k:
tmp[-1] = (tmp[-1].transpose(
(1, 2, 0)) * 255).astype(np.uint8)
if 'mask' in k:
tmp[-1] = tmp[-1].astype(np.uint8).squeeze()
if 'keypoints' in k:
tmp[-1] = slc.Keypoints(tmp[-1], 3, 4)
tensor_dict[k] = tmp
else:
el = tensor_dict[k]
tensor_dict[k] = (el.numpy()).astype(np.uint8) if isinstance(
el, torch.Tensor) else el
if 'imag' in k:
tensor_dict[k] = (tensor_dict[k].transpose(
(1, 2, 0)) * 255).astype(np.uint8)
if 'mask' in k:
tensor_dict[k] = tensor_dict[k].astype(np.uint8).squeeze()
if 'keypoints' in k:
tensor_dict[k] = slc.Keypoints(tensor_dict[k], 3, 4)
assert dc == slc.DataContainer.from_dict(tensor_dict)
def test_3x3_pad_to_20x20_center_crop_3x3_shape_stayes_unchanged(img_3x3, mask_3x3):
# Setting up the data
kpts_data = np.array([[0, 0], [0, 2], [2, 2], [2, 0]]).reshape((4, 2))
kpts = slc.Keypoints(kpts_data, 3, 3)
img, mask = img_3x3, mask_3x3
dc = slc.DataContainer((img, mask, kpts,), "IMP")
stream = slc.Stream([slt.Pad((20, 20)), slt.Crop((3, 3))])
res = stream(dc, return_torch=False)
assert (res[0][0].shape[0] == 3) and (res[0][0].shape[1] == 3)
assert (res[1][0].shape[0] == 3) and (res[1][0].shape[1] == 3)
assert (res[2][0].height == 3) and (res[2][0].width == 3)
def test_6x6_pad_to_20x20_center_crop_6x6_kpts_img(img):
# Setting up the data
kpts_data = np.array([[0, 0], [0, 5], [1, 3], [2, 0]]).reshape((4, 2))
kpts = slc.Keypoints(kpts_data, frame=(6, 6))
dc = slc.DataContainer((kpts, img), "PI")
stream = slc.Stream([slt.Pad((20, 20)), slt.Crop((6, 6))])
res = stream(dc, return_torch=False)
assert (res[1][0].shape[0] == 6) and (res[1][0].shape[1] == 6)
assert (res[0][0].frame[0] == 6) and (res[0][0].frame[1] == 6)
assert np.array_equal(res[1][0], img)
assert np.array_equal(res[0][0].data, kpts_data)
def test_pad_to_20x20_img_mask_keypoints_3x3_kpts_first(img_3x3, mask_3x3):
# Setting up the data
kpts_data = np.array([[0, 0], [0, 2], [2, 2], [2, 0]]).reshape((4, 2))
kpts = slc.Keypoints(kpts_data, 3, 3)
img, mask = img_3x3, mask_3x3
dc = slc.DataContainer((kpts, img, mask), "PIM")
transf = slt.Pad((20, 20))
res = transf(dc)
assert (res[2][0].shape[0] == 20) and (res[2][0].shape[1] == 20)
assert (res[1][0].shape[0] == 20) and (res[1][0].shape[1] == 20)
assert (res[0][0].height == 20) and (res[0][0].width == 20)
assert np.array_equal(res[0][0].data, np.array([[8, 8], [8, 10], [10, 10], [10, 8]]).reshape((4, 2)))
def test_selective_pipeline_selects_transforms_and_does_the_fusion():
ppl = slc.SelectiveStream([
slt.Rotate(angle_range=(90, 90), p=1),
slt.Rotate(angle_range=(-90, -90), p=1),
],
n=2,
probs=[0.5, 0.5],
optimize_stack=True)
kpts_data = np.array([[0, 0], [0, 1], [1, 0], [1, 1]]).reshape((4, 2))
kpts = slc.Keypoints(kpts_data, 3, 4)
dc = slc.DataContainer(kpts, 'P')
dc_res = ppl(dc, return_torch=False)
assert np.array_equal(np.eye(3),
ppl.transforms[0].state_dict['transform_matrix'])
def test_pad_to_20x20_img_mask_keypoints_3x3(img, mask):
# Setting up the data
kpts_data = np.array([[0, 0], [0, 2], [2, 2], [2, 0]]).reshape((4, 2))
kpts = slc.Keypoints(kpts_data, frame=(3, 3))
dc = slc.DataContainer((
img,
mask,
kpts,
), "IMP")
transf = slt.Pad((20, 20))
res = transf(dc)
assert (res[0][0].shape[0] == 20) and (res[0][0].shape[1] == 20)
assert (res[1][0].shape[0] == 20) and (res[1][0].shape[1] == 20)
assert (res[2][0].frame[0] == 20) and (res[2][0].frame[1] == 20)
assert np.array_equal(
res[2][0].data,
np.array([[8, 8], [8, 10], [10, 10], [10, 8]]).reshape((4, 2)))
def test_img_mask__kptsvertical_horizontal_flip_negative_axes(img, mask):
kpts_data = np.array([[0, 0], [0, 1], [1, 0], [1, 1]]).reshape((4, 2))
kpts = slc.Keypoints(kpts_data.copy(), frame=(3, 4))
dc = slc.DataContainer((img, mask, kpts), "IMP")
stream = slt.Flip(p=1, axis=-1)
dc = stream(dc)
img_res, _, _ = dc[0]
mask_res, _, _ = dc[1]
kpts_res, _, _ = dc[2]
h, w = mask.shape
assert np.array_equal(
cv2.flip(cv2.flip(img, 0), 1).reshape(h, w, 1), img_res)
assert np.array_equal(cv2.flip(cv2.flip(mask, 0), 1), mask_res)
kpts_data[:, 0] = 4 - 1 - kpts_data[:, 0]
kpts_data[:, 1] = 3 - 1 - kpts_data[:, 1]
assert np.array_equal(kpts_data, kpts_res.data)
def test_keypoint_raises_with_pts_no_frame():
with pytest.raises(ValueError):
slc.Keypoints(pts=np.asarray([[0, 1], [1, 0]]))
def test_create_empty_keypoints():
kpts = slc.Keypoints()
assert kpts.height is None
assert kpts.width is None
assert kpts.data is None
def generate_data_container_based_on_presence(img, mask, kpts_data, order,
presence):
kpts = slc.Keypoints(kpts_data.copy(), 3, 4)
n_obj1, n_obj2, n_obj3, n_obj4, n_obj5, n_obj6, n_obj7, n_obj8 = presence
dc_content = []
dc_format = ''
order = list(order)
d = {}
if n_obj1:
d['image'] = img
dc_content.append(img)
dc_format += 'I'
else:
del order[order.index('image')]
if n_obj2:
d['images'] = [img.copy() for _ in range(n_obj2)]
dc_content.extend(d['images'])
dc_format += 'I' * n_obj2
else:
del order[order.index('images')]
if n_obj3:
d['mask'] = mask
dc_content.append(mask)
dc_format += 'M'
else:
del order[order.index('mask')]
if n_obj4:
d['masks'] = [mask.copy() for i in range(n_obj4)]
dc_content.extend(d['masks'])
dc_format += 'M' * n_obj4
else:
del order[order.index('masks')]
if n_obj5:
d['keypoints'] = slc.Keypoints(kpts_data.copy(), 3, 4)
dc_content.append(kpts)
dc_format += 'P'
else:
del order[order.index('keypoints')]
if n_obj6:
d['keypoints_array'] = [
slc.Keypoints(kpts_data.copy(), 3, 4) for _ in range(n_obj6)
]
dc_content.extend(d['keypoints_array'])
dc_format += 'P' * n_obj6
else:
del order[order.index('keypoints_array')]
if n_obj7:
d['label'] = 1
dc_content.append(1)
dc_format += 'L'
else:
del order[order.index('label')]
if n_obj8:
d['labels'] = [1 for _ in range(n_obj8)]
dc_content.extend([1 for _ in range(n_obj8)])
dc_format += 'L' * n_obj8
else:
del order[order.index('labels')]
dc = slc.DataContainer(tuple(dc_content), dc_format)
reordered_d = {k: d[k] for k in order}
# This tests whether the creation from dict works as expected
dc_reordered = slc.DataContainer.from_dict(reordered_d)
return dc, dc_reordered
