kd.data.py.Resize

class kauldron.data.py.Resize(*, key: typing.Annotated[typing.Any, <object object at 0x76412092fb90>] | typing.Sequence[typing.Annotated[typing.Any, <object object at 0x76412092fb90>]] | dict[typing.Annotated[typing.Any, <object object at 0x76412092fb90>], typing.Annotated[typing.Any, <object object at 0x76412092fb90>]], size: tuple[int, int] | None = None, min_size: int | None = None, max_size: int | None = None, method: str | jax._src.image.scale.ResizeMethod | tensorflow.python.ops.image_ops_impl.ResizeMethod | None = None, antialias: bool = True)

Bases: kauldron.data.transforms.base.ElementWiseTransform

Resizes an image.

At most one of size, min_size, and max_size must be set.

size

The new size of the image. If set, the image is rescaled so that the new size matches size.

Type:

tuple[int, int] | None

min_size

The minimum size of the image. If set, the image is rescaled so that the smaller edge matches min_size.

Type:

int | None

max_size

The maximum size of the image. If set, the image is rescaled so that the larger edge matches max_size.

Type:

int | None

method

The resizing method. If None, uses area for float TF inputs, bilinear for float JAX inputs, and nearest for int inputs.

Type:

str | jax._src.image.scale.ResizeMethod | tensorflow.python.ops.image_ops_impl.ResizeMethod | None

antialias

Whether to use an anti-aliasing filter.

Type:

bool

size: tuple[int, int] | None = None

min_size: int | None = None

max_size: int | None = None

method: str | jax._src.image.scale.ResizeMethod | tensorflow.python.ops.image_ops_impl.ResizeMethod | None = None

antialias: bool = True

map_element(

element: jaxtyping.Shaped[Tensor, '*b h w c'] | jaxtyping.Shaped[ndarray, '*b h w c'] | jaxtyping.Shaped[Array, '*b h w c'],

) → jaxtyping.Shaped[Tensor, '*b h2 w2 c'] | jaxtyping.Shaped[ndarray, '*b h2 w2 c'] | jaxtyping.Shaped[Array, '*b h2 w2 c']

key: kontext.Key | Sequence[kontext.Key] | dict[kontext.Key, kontext.Key]