CropTight

class deeptrack.optical.augmentations.CropTight(eps: float | Callable[[...], float] = 1e-10, **kwargs: Any)

Bases: Augmentation

Crop an array to remove empty space.

Removes leading and trailing indices along each axis where all values are below a threshold eps. This effectively crops the array to the smallest bounding box containing values larger than eps. Currently only supports 3D arrays of shape (H, W, Z).

Parameters

eps: PropertyLike[float], optional

Threshold below which values are considered empty. Defaults to 1e-10.

Methods

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

Applies tight cropping to a NumPy array.

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

Applies tight cropping to a PyTorch tensor.

Examples

>>> import deeptrack as dt
>>> particle = dt.PointParticle(position=(32, 32))
>>> optics = dt.Fluorescence()
>>> crop_tight = dt.CropTight(eps=1e-5)
>>> image = optics(particle) >> crop_tight
>>> image.plot()