visionsim.interpolate package

Submodules

visionsim.interpolate.pose module

class visionsim.interpolate.pose.pose_interp(transforms: npt.ArrayLike, ts: TypeAliasForwardRef('npt.ArrayLike') | None = None, k=3, normalize=False)

Bases: object

Linearly interpolate between 4x4 (or 3x4) transformation matrices by interpolating it’s components

__init__(transforms: npt.ArrayLike, ts: TypeAliasForwardRef('npt.ArrayLike') | None = None, k=3, normalize=False) → None

Create a spline from the rotational and translational components of each pose. Specifically, we use scipy’s RotationSpline and BSpline respectively.

Parameters:

• transforms (npt.ArrayLike) – Poses to interpolate as matrices.

• ts (npt.ArrayLike | None, optional) – Time of each pose. Defaults to None, meaning linspace between [0, 1].

• k (int, optional) – B-spline degree for translation. Default is cubic (i.e: 3).

• normalize (bool, optional) – If true, normalize rotations by their determinants. Defaults to False.

Raises:

RuntimeError – raised when rotations are not normalized.

__call__(ts: npt.ArrayLike, order: Literal[0, 1, 2] = 0) → npt.NDArray

Compute interpolated poses, or their derivatives.

Parameters:

• ts (npt.ArrayLike) – Times of interest.

• order (Literal[0, 1, 2], optional) –

  Order of differentiation:

  • 0 (default): return pose as 4x4 matrices

  • 1: return velocities, where the first row is the angular rates

    in rad/sec and second row are positional velocities.

  • 2: return the accelerations, packaged as angular acceleration

    (in rad/sec/sec) then positional.

  Defaults to 0.

Returns:

Interpolated poses (Tx4x4) or their derivatives (2xTx3)

Return type:

npt.NDArray

Module contents

visionsim.interpolate.interpolate_poses(transforms, normalize: bool = False, n: int = 2, k: int = 3)

Interpolate between pose matrices

Parameters:

• transforms – List of pose matrices to interpolate between

• normalize (bool) – Whether the interpolation should be normalized or not

• n (int) – Number of poses to interpolate between existing poses

• k (int) – Order of spline interpolation, see pose_interp

Returns:

List of interpolated poses

visionsim.interpolate.interpolate_frames(input_dir: str | PathLike, output_dir: str | PathLike, interpolation_method: Literal['rife'] = 'rife', n: int = 2)

Interpolate between image frames

Note: Currently only RIFE is supported for frame interpolation but we intend to add more

Parameters:

• input_dir (pathlib.Path) – Path to directory containing images to interpolate

• output_dir (pathlib.Path) – Path to directory to output interpolated frames

• interpolation_method (str) – The image interpolation method used

• n (int) – Number of frames to interpolate between existing frames

visionsim.interpolate.poses_and_frames_to_json(transforms, new_poses, output_dir: str | PathLike, file_name: str = 'transforms.json')

Combines interpolated poses (matrices) and frames into a new transforms.json file

Parameters:

• transforms – Original transforms JSON file

• new_poses – List of interpolated pose matrices

• output_dir (pathlib.Path) – Path to directory containing interpolated frames

• file_name – Name of new transforms JSON file