def test_invalid_box_data(self):
with self.assertRaises(ValueError):
np_box_list.BoxList3d(length=np.array([0]),
height=np.array([0]),
width=np.array([0]),
center_x=np.array([1]),
center_y=np.array([1]),
center_z=np.array([1]),
rotation_z_radians=np.array([0]))
with self.assertRaises(ValueError):
np_box_list.BoxList3d(length=np.array([1, -1]),
height=np.array([1, 1]),
width=np.array([1, 1]),
center_x=np.array([1, 1]),
center_y=np.array([1, 1]),
center_z=np.array([1, 1]),
rotation_z_radians=np.array([0, 0]))
with self.assertRaises(ValueError):
np_box_list.BoxList3d(length=np.array([1, 1]),
height=np.array([1, 1]),
width=np.array([1, 1]),
center_x=np.array([1]),
center_y=np.array([1, 1]),
center_z=np.array([1, 1]),
rotation_z_radians=np.array([0, 0]))
def setUp(self):
boxes1 = np.array(
[[_degree_to_radians(0.0), 1.0, 4.0, 3.0, 2.0, 5.0, 5.0],
[_degree_to_radians(0.0), 2.0, 5.0, 6.0, 2.0, 7.5, 7.0]],
dtype=float)
(rotation_z_radians1, length1, height1, width1, center_x1, center_y1,
center_z1) = _decompose_box_tensor(boxes1)
boxes2 = np.array(
[[_degree_to_radians(0.0), 1.0, 3.0, 4.0, 1.0, 4.5, 8.0],
[_degree_to_radians(0.0), 14.0, 14.0, 14.0, 15.0, 8.0, 15.0],
[_degree_to_radians(0.0), 1.0, 4.0, 3.0, 2.0, 5.0, 5.0],
[_degree_to_radians(0.0), 2.0, 5.0, 6.0, 2.0, 7.5, 7.0]],
dtype=float)
(rotation_z_radians2, length2, height2, width2, center_x2, center_y2,
center_z2) = _decompose_box_tensor(boxes2)
rotation_matrix1 = np.tile(np.expand_dims(np.eye(3), axis=0),
[2, 1, 1])
rotation_matrix2 = np.tile(np.expand_dims(np.eye(3), axis=0),
[4, 1, 1])
self.boxlist1 = np_box_list.BoxList3d(
length=length1,
height=height1,
width=width1,
center_x=center_x1,
center_y=center_y1,
center_z=center_z1,
rotation_z_radians=rotation_z_radians1)
self.boxlist2 = np_box_list.BoxList3d(
length=length2,
height=height2,
width=width2,
center_x=center_x2,
center_y=center_y2,
center_z=center_z2,
rotation_z_radians=rotation_z_radians2)
self.boxlist1_matrix = np_box_list.BoxList3d(
length=length1,
height=height1,
width=width1,
center_x=center_x1,
center_y=center_y1,
center_z=center_z1,
rotation_matrix=rotation_matrix1)
self.boxlist2_matrix = np_box_list.BoxList3d(
length=length2,
height=height2,
width=width2,
center_x=center_x2,
center_y=center_y2,
center_z=center_z2,
rotation_matrix=rotation_matrix2)
def setUp(self):
boxes = np.array(
[[_degree_to_radians(0.1), 1.0, 3.0, 4.0, 1.0, 6.0, 8.0],
[_degree_to_radians(0.2), 14.0, 14.0, 14.0, 15.0, 15.0, 15.0],
[_degree_to_radians(0.1), 1.0, 4.0, 3.0, 2.0, 7.0, 5.0],
[_degree_to_radians(0.2), 2.0, 5.0, 6.0, 2.0, 10.0, 7.0]],
dtype=float)
(rotation_z_radians, length, height, width, center_x, center_y,
center_z) = _decompose_box_tensor(boxes)
self.boxlist = np_box_list.BoxList3d(
length=length,
height=height,
width=width,
center_x=center_x,
center_y=center_y,
center_z=center_z,
rotation_z_radians=rotation_z_radians)
self.boxlist.add_field('scores',
np.array([0.5, 0.7, 0.9, 0.4], dtype=float))
self.boxlist.add_field(
'labels',
np.array([[0, 0, 0, 1, 0],
[0, 1, 0, 0, 0],
[0, 0, 0, 0, 1],
[0, 0, 1, 0, 0]],
dtype=int))
def test_concatenate_boxes3d(self):
boxes1 = np.array([[0.25, 0.25, 0.75, 0.75, 1.0, 2.0, 3.0],
[0.0, 0.1, 0.5, 0.75, 0.0, 0.0, 1.0]],
dtype=np.float32)
(rotation_z_radians1, length1, height1, width1, center_x1, center_y1,
center_z1) = _decompose_box_tensor(boxes1)
boxlist1 = np_box_list.BoxList3d(
length=length1,
height=height1,
width=width1,
center_x=center_x1,
center_y=center_y1,
center_z=center_z1,
rotation_z_radians=rotation_z_radians1)
boxes2 = np.array([[0.5, 0.25, 1.0, 1.0, 2.0, 3.0, 4.0],
[0.0, 0.1, 1.0, 1.0, 10.0, 20.0, 30.0]],
dtype=np.float32)
(rotation_z_radians2, length2, height2, width2, center_x2, center_y2,
center_z2) = _decompose_box_tensor(boxes2)
boxlist2 = np_box_list.BoxList3d(
length=length2,
height=height2,
width=width2,
center_x=center_x2,
center_y=center_y2,
center_z=center_z2,
rotation_z_radians=rotation_z_radians2)
boxlists = [boxlist1, boxlist2]
boxlist_concatenated = np_box_list_ops.concatenate_boxes3d(boxlists)
boxes_concatenated_expected = np.array(
[[0.25, 0.25, 0.75, 0.75, 1.0, 2.0, 3.0],
[0.0, 0.1, 0.5, 0.75, 0.0, 0.0, 1.0],
[0.5, 0.25, 1.0, 1.0, 2.0, 3.0, 4.0],
[0.0, 0.1, 1.0, 1.0, 10.0, 20.0, 30.0]],
dtype=np.float32)
self._compare_boxlist_with_boxes(boxlist_concatenated,
boxes_concatenated_expected)
def np_nms(np_boxes_length, np_boxes_height, np_boxes_width, np_boxes_center,
np_boxes_rotation_matrix, np_boxes_score, score_thresh, iou_thresh,
max_output_size):
"""Non maximum suppression.
Args:
np_boxes_length: A numpy array of size [N, 1].
np_boxes_height: A numpy array of size [N, 1].
np_boxes_width: A numpy array of size [N, 1].
np_boxes_center: A numpy array of size [N, 3].
np_boxes_rotation_matrix: A numpy array of size [N, 3, 3].
np_boxes_score: A numpy array of size [N, num_classes].
score_thresh: scalar threshold for score (low scoring boxes are removed).
iou_thresh: scalar threshold for IOU (boxes that that high IOU overlap
with previously selected boxes are removed).
max_output_size: maximum number of retained boxes per class.
Returns:
np_nms_boxes_length: A numpy array of size [N', 1].
np_nms_boxes_height: A numpy array of size [N', 1].
np_nms_boxes_width: A numpy array of size [N', 1].
np_nms_boxes_center: A numpy array of size [N', 3].
np_nms_boxes_rotation_matrix: A numpy array of size [N', 3, 3].
np_nms_boxes_class: A numpy array of size [N', 1].
np_nms_boxes_score: A numpy array of size [N', 1].
"""
boxlist = np_box_list.BoxList3d(length=np.squeeze(np_boxes_length, axis=1),
height=np.squeeze(np_boxes_height, axis=1),
width=np.squeeze(np_boxes_width, axis=1),
center_x=np_boxes_center[:, 0],
center_y=np_boxes_center[:, 1],
center_z=np_boxes_center[:, 2],
rotation_matrix=np.reshape(
np_boxes_rotation_matrix, [-1, 3, 3]))
boxlist.add_field('scores', np_boxes_score)
boxlist_nms = np_box_list_ops.multi_class_non_max_suppression3d(
boxlist=boxlist,
score_thresh=score_thresh,
iou_thresh=iou_thresh,
max_output_size=max_output_size)
return (np.expand_dims(boxlist_nms.get_length(), axis=1),
np.expand_dims(boxlist_nms.get_height(), axis=1),
np.expand_dims(boxlist_nms.get_width(),
axis=1), boxlist_nms.get_center(),
np.expand_dims(boxlist_nms.get_rotation_matrix(), axis=1),
np.expand_dims(boxlist_nms.get_field('classes'),
axis=1).astype(np.int32),
np.expand_dims(boxlist_nms.get_field('scores'), axis=1))
def copy_boxlist(boxlist, indices=None):
"""Copy the boxes of a BoxList3d object into a new BoxList3d object.
Args:
boxlist: A np_box_list.BoxList3d object.
indices: A indices of the boxes to be copied. It is not used if None.
Returns:
new_boxlist: A new np_box_list.BoxList3d object.
"""
length = boxlist.get_length()
height = boxlist.get_height()
width = boxlist.get_width()
center_x = boxlist.get_center_x()
center_y = boxlist.get_center_y()
center_z = boxlist.get_center_z()
rotation_matrix = boxlist.get_rotation_matrix()
rotation_z_radians = boxlist.get_rotation_z_radians()
if indices is not None:
length = length[indices]
height = height[indices]
width = width[indices]
center_x = center_x[indices]
center_y = center_y[indices]
center_z = center_z[indices]
if rotation_matrix is not None:
rotation_matrix = rotation_matrix[indices, :, :]
if rotation_z_radians is not None:
rotation_z_radians = rotation_z_radians[indices]
new_boxlist = np_box_list.BoxList3d(
length=length,
height=height,
width=width,
center_x=center_x,
center_y=center_y,
center_z=center_z,
rotation_matrix=rotation_matrix,
rotation_z_radians=rotation_z_radians)
return new_boxlist
def setUp(self):
boxes = np.array(
[[_degree_to_radians(0.0), 1.0, 1.0, 1.0, 0.0, 0.0, 0.0],
[_degree_to_radians(10.0), 1.0, 1.0, 1.0, 0.0, 0.0, 0.0],
[_degree_to_radians(20.0), 1.0, 1.0, 1.0, 0.0, 0.0, 0.0],
[_degree_to_radians(-10.0), 1.0, 1.0, 1.0, 0.0, 0.0, 0.0],
[_degree_to_radians(0.0), 1.0, 1.0, 1.0, 10.0, 0.0, 0.0],
[_degree_to_radians(10.0), 1.0, 1.0, 1.0, 10.0, 0.0, 0.0],
[_degree_to_radians(20.0), 1.0, 1.0, 1.0, 10.0, 0.0, 0.0]],
dtype=float)
(rotation_z_radians, length, height, width, center_x, center_y,
center_z) = _decompose_box_tensor(boxes)
self.boxlist = np_box_list.BoxList3d(
length=length,
height=height,
width=width,
center_x=center_x,
center_y=center_y,
center_z=center_z,
rotation_z_radians=rotation_z_radians)
self.boxlist.add_field(
'scores', np.array([0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7],
dtype=float))
def concatenate_boxes3d(boxlists, fields=None):
"""Concatenates list of BoxList3d-s.
This op concatenates a list of input BoxList3d-s into a larger BoxList. It
also handles concatenation of BoxList fields as long as the field tensor
shapes are equal except for the first dimension.
Args:
boxlists: list of BoxList3d objects.
fields: optional list of fields to also concatenate. By default, all
fields from the first BoxList3d in the list are included in the
concatenation.
Returns:
A BoxList3d with number of boxes equal to
sum([boxlist.num_boxes() for boxlist in boxlists]).
Raises:
ValueError: If boxlists is invalid (i.e., is not a list, is empty, or
contains non BoxList3d objects), or if requested fields are not contained
in all boxlists.
"""
if not isinstance(boxlists, list):
raise ValueError('boxlists should be a list')
if not boxlists:
raise ValueError('boxlists should have nonzero length')
for boxlist in boxlists:
if not isinstance(boxlist, np_box_list.BoxList3d):
raise ValueError(
'all elements of boxlists should be BoxList3d objects')
length_list = []
height_list = []
width_list = []
center_x_list = []
center_y_list = []
center_z_list = []
rotation_matrix_list = []
rotation_z_radians_list = []
for boxlist in boxlists:
length_list.append(boxlist.get_length())
height_list.append(boxlist.get_height())
width_list.append(boxlist.get_width())
center_x_list.append(boxlist.get_center_x())
center_y_list.append(boxlist.get_center_y())
center_z_list.append(boxlist.get_center_z())
rotation_matrix_list.append(boxlist.get_rotation_matrix())
rotation_z_radians_list.append(boxlist.get_rotation_z_radians())
length = np.concatenate(length_list)
height = np.concatenate(height_list)
width = np.concatenate(width_list)
center_x = np.concatenate(center_x_list)
center_y = np.concatenate(center_y_list)
center_z = np.concatenate(center_z_list)
if rotation_matrix_list[0] is None:
rotation_matrix = None
else:
rotation_matrix = np.concatenate(rotation_matrix_list)
if rotation_z_radians_list[0] is None:
rotation_z_radians = None
else:
rotation_z_radians = np.concatenate(rotation_z_radians_list)
concatenated = np_box_list.BoxList3d(length=length,
height=height,
width=width,
center_x=center_x,
center_y=center_y,
center_z=center_z,
rotation_matrix=rotation_matrix,
rotation_z_radians=rotation_z_radians)
if fields is None:
fields = boxlists[0].get_extra_fields()
for field in fields:
first_field_shape = boxlists[0].get_field(field).shape
first_field_shape = first_field_shape[1:]
for boxlist in boxlists:
if not boxlist.has_field(field):
raise ValueError('boxlist must contain all requested fields')
field_shape = boxlist.get_field(field).shape
field_shape = field_shape[1:]
if field_shape != first_field_shape:
raise ValueError(
'field %s must have same shape for all boxlists '
'except for the 0th dimension.' % field)
concatenated_field = np.concatenate(
[boxlist.get_field(field) for boxlist in boxlists], axis=0)
concatenated.add_field(field, concatenated_field)
return concatenated
