def init_from_edges(cls,
points,
edges,
labels_to_masks,
copy=True,
skip_checks=False):
r"""
Construct a :map:`LabelledPointUndirectedGraph` from an edges array.
See :map:`PointUndirectedGraph` for more information.
Parameters
----------
points : ``(n_vertices, n_dims, )`` `ndarray`
The array of point locations.
edges : ``(n_edges, 2, )`` `ndarray` or ``None``
The `ndarray` of edges, i.e. all the pairs of vertices that are
connected with an edge. If ``None``, then an empty adjacency
matrix is created.
labels_to_masks : `ordereddict` `{str -> bool ndarray}`
For each label, the mask that specifies the indices in to the
points that belong to the label.
copy : `bool`, optional
If ``False``, the `adjacency_matrix` will not be copied on
assignment.
skip_checks : `bool`, optional
If ``True``, no checks will be performed.
"""
adjacency_matrix = _convert_edges_to_symmetric_adjacency_matrix(
edges, points.shape[0])
return cls(points,
adjacency_matrix,
labels_to_masks,
copy=copy,
skip_checks=skip_checks)
def init_from_edges(cls, points, edges, labels_to_masks, copy=True,
skip_checks=False):
r"""
Construct a :map:`LabelledPointUndirectedGraph` from an edges array.
See :map:`PointUndirectedGraph` for more information.
Parameters
----------
points : ``(n_vertices, n_dims, )`` `ndarray`
The array of point locations.
edges : ``(n_edges, 2, )`` `ndarray` or ``None``
The `ndarray` of edges, i.e. all the pairs of vertices that are
connected with an edge. If ``None``, then an empty adjacency
matrix is created.
labels_to_masks : `ordereddict` `{str -> bool ndarray}`
For each label, the mask that specifies the indices in to the
points that belong to the label.
copy : `bool`, optional
If ``False``, the `adjacency_matrix` will not be copied on
assignment.
skip_checks : `bool`, optional
If ``True``, no checks will be performed.
"""
adjacency_matrix = _convert_edges_to_symmetric_adjacency_matrix(
edges, points.shape[0])
return cls(points, adjacency_matrix, labels_to_masks, copy=copy,
skip_checks=skip_checks)
def __setstate__(self, state_dict):
# TODO: Deprecate this - this handles importing old-style LandmarkGroup
if '_pointcloud' in state_dict:
from menpo.base import MenpoDeprecationWarning
warnings.warn('menpo.landmark.LandmarkGroup is now deprecated and '
'has been moved to menpo.shape.LandmarkGroup.',
MenpoDeprecationWarning)
_pointcloud = state_dict.pop('_pointcloud')
state_dict['points'] = _pointcloud.points
if type(_pointcloud) == PointCloud:
adj_mat = _convert_edges_to_symmetric_adjacency_matrix(
[], _pointcloud.shape[0])
elif isinstance(_pointcloud, PointGraph):
a = _pointcloud.adjacency_matrix
# Ensure that the matrix is symmetric
adj_mat = a.maximum(a.T)
elif isinstance(_pointcloud, TriMesh):
warnings.warn('menpo.landmark.LandmarkGroup is now deprecated.'
'The underlying ._pointcloud was a '
'menpo.shape.TriMesh and this has been cast down '
'to an UndirectedPointGraph subclass.')
adj_mat = _pointcloud.as_pointgraph(copy=False).adjacency_matrix
else:
raise ValueError('Unexpected PointCloud type ({})'.format(
type(_pointcloud)))
state_dict['adjacency_matrix'] = adj_mat
self.__dict__.update(state_dict)
def init_from_indices_mapping(cls,
points,
adjacency,
labels_to_indices,
copy=True):
r"""
Static constructor to create a :map:`LabelledPointUndirectedGraph` from
an ordered dictionary that maps a set of indices .
Parameters
----------
points : :map:`PointCloud`
The points representing the landmarks.
adjacency : ``(n_vertices, n_vertices, )`` `ndarray`, `csr_matrix` or `list` of edges
The adjacency matrix of the graph, or a list of edges representing
adjacency.
labels_to_indices : `ordereddict` {`str` -> `int ndarray`}
For each label, the indices in to the points that belong to the
label.
copy : `boolean`, optional
If ``True``, a copy of the data is stored on the group.
Returns
-------
labelled_pointgraph : :map:`LabelledPointUndirectedGraph`
Labelled point undirected graph wrapping the given points with the
given semantic labels applied.
Raises
------
ValueError
If `dict` passed instead of `OrderedDict`
ValueError
If any of the label masks differs in size to the points.
ValueError
If there exists any point in the points that is not covered
by a label.
"""
adjacency = np.array(adjacency)
if adjacency.shape[0] != adjacency.shape[1] and adjacency.shape[1] == 2:
adjacency = _convert_edges_to_symmetric_adjacency_matrix(
adjacency, points.shape[0])
labels_to_masks = indices_to_masks(labels_to_indices, points.shape[0])
return LabelledPointUndirectedGraph(points,
adjacency,
labels_to_masks,
copy=copy)
def init_from_indices_mapping(cls, points, adjacency,
labels_to_indices, copy=True):
r"""
Static constructor to create a :map:`LabelledPointUndirectedGraph` from
an ordered dictionary that maps a set of indices .
Parameters
----------
points : :map:`PointCloud`
The points representing the landmarks.
adjacency : ``(n_vertices, n_vertices, )`` `ndarray`, `csr_matrix` or `list` of edges
The adjacency matrix of the graph, or a list of edges representing
adjacency.
labels_to_indices : `ordereddict` {`str` -> `int ndarray`}
For each label, the indices in to the points that belong to the
label.
copy : `boolean`, optional
If ``True``, a copy of the data is stored on the group.
Returns
-------
labelled_pointgraph : :map:`LabelledPointUndirectedGraph`
Labelled point undirected graph wrapping the given points with the
given semantic labels applied.
Raises
------
ValueError
If `dict` passed instead of `OrderedDict`
ValueError
If any of the label masks differs in size to the points.
ValueError
If there exists any point in the points that is not covered
by a label.
"""
adjacency = np.array(adjacency)
if adjacency.shape[0] != adjacency.shape[1] and adjacency.shape[1] == 2:
adjacency = _convert_edges_to_symmetric_adjacency_matrix(
adjacency, points.shape[0])
labels_to_masks = indices_to_masks(labels_to_indices,
points.shape[0])
return LabelledPointUndirectedGraph(points, adjacency, labels_to_masks,
copy=copy)
def __setstate__(self, state_dict):
# TODO: Deprecate this - this handles importing old-style LandmarkGroup
if "_pointcloud" in state_dict:
from menpo.base import MenpoDeprecationWarning
warnings.warn(
"menpo.landmark.LandmarkGroup is now deprecated and "
"has been moved to menpo.shape.LandmarkGroup.",
MenpoDeprecationWarning,
)
_pointcloud = state_dict.pop("_pointcloud")
state_dict["points"] = _pointcloud.points
# the shape on old landmarks *itself* was allowed to have landmarks
# (of course it was very frequently None though, see
# https://github.com/menpo/menpo/blob/v0.7.7/menpo/landmark/base.py#L24)
# In the new word, self has the same behavior, so move the
# landmarks across here.
# In the vast majority of cases, this will simply be None.
state_dict["_landmarks"] = _pointcloud._landmarks
if type(_pointcloud) == PointCloud:
adj_mat = _convert_edges_to_symmetric_adjacency_matrix(
[], _pointcloud.n_points)
elif isinstance(_pointcloud, PointGraph):
a = _pointcloud.adjacency_matrix
# Ensure that the matrix is symmetric
adj_mat = a.maximum(a.T)
elif isinstance(_pointcloud, TriMesh):
warnings.warn(
"menpo.landmark.LandmarkGroup is now deprecated."
"The underlying ._pointcloud was a "
"menpo.shape.TriMesh and this has been cast down "
"to an UndirectedPointGraph subclass.")
adj_mat = _pointcloud.as_pointgraph(
copy=False).adjacency_matrix
else:
raise ValueError("Unexpected PointCloud type ({})".format(
type(_pointcloud)))
state_dict["adjacency_matrix"] = adj_mat
self.__dict__.update(state_dict)
def __setstate__(self, state_dict):
# TODO: Deprecate this - this handles importing old-style LandmarkGroup
if '_pointcloud' in state_dict:
from menpo.base import MenpoDeprecationWarning
warnings.warn('menpo.landmark.LandmarkGroup is now deprecated and '
'has been moved to menpo.shape.LandmarkGroup.',
MenpoDeprecationWarning)
_pointcloud = state_dict.pop('_pointcloud')
state_dict['points'] = _pointcloud.points
# the shape on old landmarks *itself* was allowed to have landmarks
# (of course it was very frequently None though, see
# https://github.com/menpo/menpo/blob/v0.7.7/menpo/landmark/base.py#L24)
# In the new word, self has the same behavior, so move the
# landmarks across here.
# In the vast majority of cases, this will simply be None.
state_dict['_landmarks'] = _pointcloud._landmarks
if type(_pointcloud) == PointCloud:
adj_mat = _convert_edges_to_symmetric_adjacency_matrix(
[], _pointcloud.n_points)
elif isinstance(_pointcloud, PointGraph):
a = _pointcloud.adjacency_matrix
# Ensure that the matrix is symmetric
adj_mat = a.maximum(a.T)
elif isinstance(_pointcloud, TriMesh):
warnings.warn('menpo.landmark.LandmarkGroup is now deprecated.'
'The underlying ._pointcloud was a '
'menpo.shape.TriMesh and this has been cast down '
'to an UndirectedPointGraph subclass.')
adj_mat = _pointcloud.as_pointgraph(copy=False).adjacency_matrix
else:
raise ValueError('Unexpected PointCloud type ({})'.format(
type(_pointcloud)))
state_dict['adjacency_matrix'] = adj_mat
self.__dict__.update(state_dict)