AffineMatrix

voxel.AffineMatrix(data=None, device=None, dtype=torch.float32)

Affine matrix (4x4) transform for a 3D coordinate system.

Parameters:

• data (Tensor, default: None ) –

  A 3x3, 3x4, or 4x4 tensor. If None, the matrix is initialized with the identity.

• device (device, default: None ) –

  Device of the constructed matrix.

• dtype (dtype, default: float32 ) –

  Data type of the constructed matrix. Defaults to float32. Use float64 to retain full-precision.

tensor: torch.Tensor property

Tensor data of shape (4, 4).

save(filename, fmt=None) -> None

Save the affine matrix to a file.

Parameters:

• filename (PathLike) –

  The path to the file to save.

• fmt (str, default: None ) –

  The format of the file. If None, the format is determined by the file extension.

detach() -> T

Detach the matrix tensor from the current computational graph.

Returns:

• T –

  A new affine with the detached matrix tensor.

to(device) -> T

Move the matrix tensor to a device.

Parameters:

• device (device) –

  The target device.

Returns:

• T –

  A new affine with the matrix tensor on the target device.

cuda() -> T

Move the matrix tensor to the GPU.

Returns:

• T –

  A new affine with the matrix tensor on the GPU.

cpu() -> T

Move the matrix tensor to the CPU.

Returns:

• T –

  A new affine with the matrix tensor on the CPU.

inverse() -> AffineMatrix

Invert the matrix.

Returns:

• AffineMatrix –

  Inverted affine matrix.

map(coords) -> torch.Tensor

Map a set of 3D coordinates through the matrix transform.

Parameters:

• coords (Tensor) –

  A tensor of coordinates with shape (..., 3).

Returns:

• Tensor –

  Mapped coordinates with the same shape as the input.

transform(coords) -> torch.Tensor

Deprecated alias of map.