def test_single_list_patch():
patch_shape = (21, 7)
n_channels = 4
im = Image.init_blank(patch_shape, n_channels)
patch = [Image(np.ones((n_channels,) + patch_shape)),
Image(2 * np.ones((n_channels,) + patch_shape))]
patch_center = PointCloud(np.array([[10., 3.], [11., 3.]]))
im.set_patches(patch, patch_center, offset=(0, 0), offset_index=0)
res = np.ones(patch_shape)
res[1:-1, :] = 2
assert_array_equal(im.pixels[2, ...], res)
def test_set_patches_around_landmarks():
patch_shape = (21, 12)
image = mio.import_builtin_asset.lenna_png()
patches1 = image.extract_patches_around_landmarks(
patch_shape=patch_shape, as_single_array=True)
new_image1 = Image.init_blank(image.shape, image.n_channels)
new_image1.landmarks['LJSON'] = image.landmarks['LJSON']
new_image1.set_patches_around_landmarks(patches1)
patches2 = image.extract_patches_around_landmarks(
patch_shape=patch_shape, as_single_array=False)
new_image2 = Image.init_blank(image.shape, image.n_channels)
new_image2.landmarks['LJSON'] = image.landmarks['LJSON']
new_image2.set_patches_around_landmarks(patches2)
assert_array_equal(new_image1.pixels, new_image2.pixels)
def test_set_patches_around_landmarks():
patch_shape = (21, 12)
image = mio.import_builtin_asset.takeo_ppm()
patches1 = image.extract_patches_around_landmarks(patch_shape=patch_shape,
as_single_array=True)
new_image1 = Image.init_blank(image.shape, image.n_channels)
new_image1.landmarks["PTS"] = image.landmarks["PTS"]
extracted1 = new_image1.set_patches_around_landmarks(patches1)
patches2 = image.extract_patches_around_landmarks(patch_shape=patch_shape,
as_single_array=False)
new_image2 = Image.init_blank(image.shape, image.n_channels)
new_image2.landmarks["PTS"] = image.landmarks["PTS"]
extracted2 = new_image2.set_patches_around_landmarks(patches2)
assert_array_equal(extracted1.pixels, extracted2.pixels)
def test_set_patches_around_landmarks():
patch_shape = (21, 12)
image = mio.import_builtin_asset.takeo_ppm()
patches1 = image.extract_patches_around_landmarks(
patch_shape=patch_shape, as_single_array=True)
new_image1 = Image.init_blank(image.shape, image.n_channels)
new_image1.landmarks['PTS'] = image.landmarks['PTS']
extracted1 = new_image1.set_patches_around_landmarks(patches1)
patches2 = image.extract_patches_around_landmarks(
patch_shape=patch_shape, as_single_array=False)
new_image2 = Image.init_blank(image.shape, image.n_channels)
new_image2.landmarks['PTS'] = image.landmarks['PTS']
extracted2 = new_image2.set_patches_around_landmarks(patches2)
assert_array_equal(extracted1.pixels, extracted2.pixels)
def test_single_list_patch():
patch_shape = (8, 7)
n_channels = 4
im = Image.init_blank((32, 32), n_channels)
patch = [
Image(np.full((n_channels,) + patch_shape, 1)),
Image(np.full((n_channels,) + patch_shape, 2)), # Should be unused
Image(np.full((n_channels,) + patch_shape, 3)),
Image(np.full((n_channels,) + patch_shape, 4)),
] # Should be unused
patch_center = PointCloud(np.array([[4.0, 4.0], [16.0, 16.0]]))
new_im = im.set_patches(patch, patch_center, offset_index=0)
res = np.zeros((32, 32))
res[:8, 1:8] = 1
res[12:20, 13:20] = 3
assert_array_equal(new_im.pixels[0], res)
def landmark_template(mesh, img_shape=(320, 240), verbose=False):
fitter = load_balanced_frontal_face_fitter()
detector = load_dlib_frontal_face_detector()
camera = perspective_camera_for_template(img_shape)
# Pre-process - align the mesh roughly with the template
aligned_mesh = prepare_template_reference_space(mesh)
mesh_in_img = camera.apply(aligned_mesh)
bcs = rasterize_barycentric_coordinate_images(mesh_in_img, img_shape)
img = rasterize_mesh_from_barycentric_coordinate_images(mesh_in_img, *bcs)
shape_img = rasterize_shape_image_from_barycentric_coordinate_images(
mesh, *bcs)
# 2. Find the one bounding box in the rendered image
bboxes = detector(img)
if len(bboxes) != 1:
raise ValueError("Expected to find one face - found {}".format(
len(bboxes)))
else:
if verbose:
print('Detected 1 face')
# 3. Fit from the bounding box
fr = fitter.fit_from_bb(img, bboxes[0])
if verbose:
print('AMM fitting successfully completed')
# 4. Sample the XYZ image to build back the landmarks
img_lms = fr.final_shape
# test to see if the landmark fell on the 3D surface or not
mesh.landmarks["ibug68"] = PointCloud(Image.sample(shape_img, img_lms).T)
mask = np.zeros(68, dtype=np.bool)
mask[30] = True
mesh.landmarks["nosetip"] = mesh.landmarks["ibug68"].lms.from_mask(mask)
def test_int_pointcloud():
image = Image.init_blank([100, 100])
patch_shape = (16, 16)
landmarks = PointCloud(np.array([[50, 50]]))
patches = image.extract_patches(
landmarks, patch_shape=patch_shape, as_single_array=False
)
assert patches[0].pixels.dtype == np.float
def test_uint16_type():
image = Image.init_blank([100, 100], dtype=np.uint16)
patch_shape = (16, 16)
landmarks = PointCloud(np.array([[50, 50.]]))
patches = image.extract_patches(landmarks,
patch_shape=patch_shape,
as_single_array=False)
assert (patches[0].pixels.dtype == np.uint16)
def test_int_pointcloud():
image = Image.init_blank([100, 100])
patch_shape = (16, 16)
landmarks = PointCloud(np.array([[50, 50]]))
patches = image.extract_patches(landmarks,
patch_shape=patch_shape,
as_single_array=False)
assert(patches[0].pixels.dtype == np.float)
def test_convert_patches_list_to_single_array():
patch_shape = (7, 2)
n_channels = 10
n_centers = 2
n_offsets = 2
patches_list = [Image(1 * np.ones((n_channels,) + patch_shape)),
Image(2 * np.ones((n_channels,) + patch_shape)),
Image(3 * np.ones((n_channels,) + patch_shape)),
Image(4 * np.ones((n_channels,) + patch_shape))]
patches_array = np.zeros((n_centers, n_offsets, n_channels) + patch_shape)
patches_array[0, 0, ...] = patches_list[0].pixels
patches_array[0, 1, ...] = patches_list[1].pixels
patches_array[1, 0, ...] = patches_list[2].pixels
patches_array[1, 1, ...] = patches_list[3].pixels
assert_array_equal(
_convert_patches_list_to_single_array(patches_list, n_centers),
patches_array)
def test_single_ndarray_patch():
patch_shape = (21, 7)
n_channels = 4
im = Image.init_blank(patch_shape, n_channels)
patch = np.zeros((2, 2, n_channels) + patch_shape)
patch[1, 0, ...] = np.ones((n_channels, ) + patch_shape)
patch[1, 1, ...] = 2 * np.ones((n_channels, ) + patch_shape)
patch_center = PointCloud(np.array([[10., 3.], [11., 3.]]))
new_im = im.set_patches(patch, patch_center, offset=(0, 0), offset_index=1)
res = np.zeros(patch_shape)
res[1:-1, :] = 2
assert_array_equal(new_im.pixels[2, ...], res)
def test_single_ndarray_patch():
patch_shape = (21, 7)
n_channels = 4
im = Image.init_blank(patch_shape, n_channels)
patch = np.zeros((2, 2, n_channels) + patch_shape)
patch[1, 0, ...] = np.ones((n_channels,) + patch_shape)
patch[1, 1, ...] = 2 * np.ones((n_channels,) + patch_shape)
patch_center = PointCloud(np.array([[10., 3.], [11., 3.]]))
new_im = im.set_patches(patch, patch_center, offset=(0, 0), offset_index=1)
res = np.zeros(patch_shape)
res[1:-1, :] = 2
assert_array_equal(new_im.pixels[2, ...], res)
def warp_to(self,
template_mask,
transform,
warp_landmarks=False,
interpolator='scipy',
**kwargs):
r"""
Warps this BooleanImage into a different reference space.
Parameters
----------
template_mask : :class:`menpo.image.boolean.BooleanImage`
Defines the shape of the result, and what pixels should be
sampled.
transform : :class:`menpo.transform.base.Transform`
Transform **from the template space back to this image**.
Defines, for each True pixel location on the template, which pixel
location should be sampled from on this image.
warp_landmarks : bool, optional
If ``True``, warped_image will have the same landmark dictionary
as self, but with each landmark updated to the warped position.
Default: ``False``
interpolator : 'scipy' or 'c', optional
The interpolator that should be used to perform the warp.
Default: 'scipy'
kwargs : dict
Passed through to the interpolator. See `menpo.interpolation`
for details.
Returns
-------
warped_image : type(self)
A copy of this image, warped.
"""
# enforce the order as 0, for this boolean data, then call super
manually_set_order = kwargs.get('order', 0)
if manually_set_order != 0:
raise ValueError(
"The order of the interpolation on a boolean image has to be "
"0 (attempted to set {})".format(manually_set_order))
kwargs['order'] = 0
return Image.warp_to(self,
template_mask,
transform,
warp_landmarks=warp_landmarks,
interpolator=interpolator,
**kwargs)
def test_single_ndarray_patch():
patch_shape = (8, 7)
n_channels = 4
im = Image.init_blank((32, 32), n_channels)
patch = np.zeros((2, 2, n_channels) + patch_shape)
patch[0, 0, ...] = np.full((n_channels,) + patch_shape, 1) # Should be unused
patch[0, 1, ...] = np.full((n_channels,) + patch_shape, 2)
patch[1, 0, ...] = np.full((n_channels,) + patch_shape, 3) # Should be unused
patch[1, 1, ...] = np.full((n_channels,) + patch_shape, 4)
patch_center = PointCloud(np.array([[4.0, 4.0], [16.0, 16.0]]))
new_im = im.set_patches(patch, patch_center, offset_index=1)
res = np.zeros((32, 32))
res[:8, 1:8] = 2
res[12:20, 13:20] = 4
assert_array_equal(new_im.pixels[2], res)
def warp_to(self, template_mask, transform, warp_landmarks=False,
interpolator='scipy', **kwargs):
r"""
Warps this BooleanImage into a different reference space.
Parameters
----------
template_mask : :class:`menpo.image.boolean.BooleanImage`
Defines the shape of the result, and what pixels should be
sampled.
transform : :class:`menpo.transform.base.Transform`
Transform **from the template space back to this image**.
Defines, for each True pixel location on the template, which pixel
location should be sampled from on this image.
warp_landmarks : bool, optional
If ``True``, warped_image will have the same landmark dictionary
as self, but with each landmark updated to the warped position.
Default: ``False``
interpolator : 'scipy' or 'c', optional
The interpolator that should be used to perform the warp.
Default: 'scipy'
kwargs : dict
Passed through to the interpolator. See `menpo.interpolation`
for details.
Returns
-------
warped_image : type(self)
A copy of this image, warped.
"""
# enforce the order as 0, for this boolean data, then call super
manually_set_order = kwargs.get('order', 0)
if manually_set_order != 0:
raise ValueError(
"The order of the interpolation on a boolean image has to be "
"0 (attempted to set {})".format(manually_set_order))
kwargs['order'] = 0
return Image.warp_to(self, template_mask, transform,
warp_landmarks=warp_landmarks,
interpolator=interpolator, **kwargs)
