Pad

class deeptrack.optical.augmentations.Pad(px: int | tuple[int, ...] | list[int] | Callable[[...], int | tuple[int, ...] | list[int]] = (0, 0, 0, 0), mode: str | Callable[[...], str] = 'constant', cval: float | Callable[[...], float] = 0, **kwargs: Any)

Bases: Augmentation

Pad an image by adding extra pixels along specified axes.

This augmentation adds padding to an image using functionality similar to numpy.pad. The padding is specified using a flat sequence px describing the number of pixels added before and after each axis.

Parameters

px: PropertyLike[int | tuple[int, …] | list[int]]

Amount of padding for each axis, specified as a flat sequence

(before_axis0, after_axis0, before_axis1, after_axis1, …)

If a single integer is provided, the same padding is applied before and after every axis.

mode: PropertyLike[str], optional

Padding mode used when extending the array. Supported modes follow numpy.pad and torch.nn.functional.pad. Defaults to “constant”.

cval: PropertyLike[float], optional

Constant value used when mode=”constant”. Defaults to 0.

Methods

_get_numpy(image, **kwargs) -> np.ndarray

Apply padding to a NumPy array.

_get_torch(image, **kwargs) -> torch.Tensor

Apply padding to a PyTorch tensor.

Returns

np.ndarray or torch.Tensor

The padded image.

Examples

>>> import deeptrack as dt
>>> particle = dt.PointParticle(position=(32, 32))
>>> optics = dt.Fluorescence()
>>> pad = dt.Pad(px=(10, 10, 5, 5), mode="constant", cval=0)
>>> image = optics(particle) >> pad
>>> print(image.resolve().shape)