Skip to content

BoundingBox

voxel.BoundingBox(center=None, rotation=None, extent=None)

An oriented 3D bounding box defined by a center, rotation, and half-length extent. Instances are immutable: all modifying operations return new boxes, and the corner points are computed lazily and cached.

Parameters:

  • center (Tensor, default: None ) –

    Box center of shape (3,). Defaults to the origin.

  • rotation (Tensor, default: None ) –

    Box rotation matrix of shape (3, 3). Defaults to the identity.

  • extent (Tensor, default: None ) –

    Non-negative box half-lengths of shape (3,). Defaults to ones.

center: torch.Tensor property

Box center coordinate of shape (3,).

rotation: torch.Tensor property

Box rotation matrix of shape (3, 3), with columns as box axes.

extent: torch.Tensor property

Box half-lengths of shape (3,) along each box axis.

device: torch.device property

Device of the box parameters.

from_min_max(min_point, max_point) -> BoundingBox classmethod

Construct an axis-aligned bounding box from lower and upper corner coordinates.

Parameters:

  • min_point (Tensor) –

    Lower corner coordinate of shape (3,).

  • max_point (Tensor) –

    Upper corner coordinate of shape (3,).

Returns:

from_points(points) -> BoundingBox classmethod

Construct an axis-aligned bounding box enclosing a point cloud.

Parameters:

  • points (Tensor) –

    Coordinate point cloud of shape (N, 3).

Returns:

to(device) -> BoundingBox

Move all bounding box parameters to a device.

Parameters:

  • device (device) –

    The target device.

Returns:

cpu() -> BoundingBox

Move all bounding box parameters to the CPU.

Returns:

cuda() -> BoundingBox

Move all bounding box parameters to the GPU.

Returns:

detach() -> BoundingBox

Detach all bounding box parameters from the current graph.

Returns:

save(filename) -> None

Save the bounding box to file.

Parameters:

  • filename (Path) –

    File destination.

corner_points() -> torch.Tensor

Compute the corner points of the bounding box. The result is cached after the first call.

Returns:

  • Tensor –

    Corner point tensor of shape (8, 3).

min_max_coords() -> tuple[torch.Tensor, torch.Tensor]

Compute the minimum and maximum coordinates of the bounding box.

Returns:

  • tuple –

    Minimum and maximum coordinates.

mesh() -> vx.Mesh

Construct a rectangular box mesh with outward-facing windings from the bounding box.

Returns:

  • Mesh –

    Rectangular box mesh.

geometry(spacing=None, *components) -> vx.AcquisitionGeometry

Construct an acquisition geometry from the bounding box.

Parameters:

  • spacing (float or Tensor, default: None ) –

    Desired voxel spacing of the geometry. Defaults to ones.

  • *components (float, default: () ) –

    Additional components of spacing, allowing values to be passed as separate positional arguments, e.g. geometry(1, 1, 2).

Returns:

shift(delta, *components) -> BoundingBox

Translate the bounding box center.

Parameters:

  • delta (float or Tensor) –

    Translation delta, as a scalar or (3,) tensor, applied in the ambient coordinate frame.

  • *components (float, default: () ) –

    Additional components of delta, allowing values to be passed as separate positional arguments, e.g. shift(1, 0, 0).

Returns:

scale(factor, *components) -> BoundingBox

Scale the bounding box extent about its center.

Parameters:

  • factor (float or Tensor) –

    Scaling factor, as a scalar or (3,) tensor of per-axis factors in the box frame.

  • *components (float, default: () ) –

    Additional components of factor, allowing values to be passed as separate positional arguments, e.g. scale(1, 1, 2).

Returns:

pad(margin, *components) -> BoundingBox

Pad (or shrink) the bounding box sides by a margin.

Parameters:

  • margin (float or Tensor) –

    Margin delta for each side of the box, in box-frame units. Can be a scalar, a (3,) tensor of per-axis margins, or a (3, 2) tensor of per-side (lower, upper) margins. Negative values shrink the box.

  • *components (float, default: () ) –

    Additional components of margin, allowing values to be passed as separate positional arguments, e.g. pad(1, 2, 3).

Returns:

trim(margin, *components) -> BoundingBox

Trim the bounding box sides by a margin. This is the inverse of pad.

Parameters:

  • margin (float or Tensor) –

    Margin delta for each side of the box, in box-frame units. Can be a scalar, a (3,) tensor of per-axis margins, or a (3, 2) tensor of per-side (lower, upper) margins.

  • *components (float, default: () ) –

    Additional components of margin, allowing values to be passed as separate positional arguments, e.g. trim(1, 2, 3).

Returns:

rotate(rotation, *components, degrees=True) -> BoundingBox

Applies a rotation to the bounding box.

Parameters:

  • rotation (Tensor) –

    Rotation angles. If degrees is True, the angles are in degrees, otherwise they are in radians.

  • *components (float, default: () ) –

    Additional components of rotation, allowing angles to be passed as separate positional arguments, e.g. rotate(0, 0, 45).

  • degrees (bool, default: True ) –

    Whether the angles are defined as degrees or, alternatively, as radians.

Returns:

transform(matrix) -> BoundingBox

Transform the bounding box through an affine matrix. This is exact as long as the box axes remain orthogonal under the transform (any rotation combined with scaling along the box axes), and an approximation under shear.

Parameters:

Returns:

fit_extent(points) -> BoundingBox

Fit the extent of the bounding box to a set of points, keeping the current rotation.

Parameters:

  • points (Tensor, BoundingBox, Mesh, or AcquisitionGeometry) –

    Coordinate point cloud of shape (N, 3), or an object defining one.

Returns:

fine_tune(points) -> BoundingBox

Orient the bounding box to minimize the volume of the box enclosing a point cloud.

Parameters:

  • points (Tensor) –

    Coordinate point cloud of shape (N, 3).

Returns: