Skip to content

module api.transforms.image.spatial

Global Variables

  • interp_codes

function Resize

Resize(
    src,
    dsize_w: int = None,
    dsize_h: int = None,
    scale: float = None,
    interpolation: str = 'nearest',
    keep_ratio: bool = False,
    return_scale: bool = False
)

Args:

  • src (np.ndarray): The input image to be resized.

  • dsize_w (int, optional): Target width. Higher priority than scale. Defaults to None.

  • dsize_h (int, optional): Target height. Should be assigned if dsize_w is specified. Defaults to None.

  • scale (float, optional): Calculate target size using this scale. Defaults to None.

  • interpolation (str, optional): Interpolation method, accepted values are "nearest", "bilinear", "bicubic", "area", "lanczos".. Defaults to "nearest".

  • keep_ratio (bool, optional): If dsize is specified, the image will be rescaled as large as possible and keep its aspect ratio, and padding redundant areas with zeros. Defaults to False.

  • return_scale (bool, optional): Whether to return actual w_scale and h_scale. Defaults to False.

Raises:

  • TypeError: when only one of dsize_w and dsize_h is specified.

  • TypeError: when scale is negative.

Returns:

  • tuple | ndarray: (resized_img, w_scale, h_scale) or resized_img

function Crop

Crop(src: ndarray, roi: Tuple)

crop a region of interest from the src array.

Args:

  • src (np.ndarray): source array (H, W) or (H, W, C)

  • roi (Tuple): region of interst (top, bottom, left, right)

Returns:

  • np.ndarray: src[top:bottom, left:right, ...]

function Flip

Flip(src: ndarray, direction: str = 'horizontal')

flip the src image.

Args:

  • src (np.ndarray): image (H, W) or (H, W, C)

  • direction (str, optional): choose in "horizontal" and "vertical". Defaults to "horizontal".

Raises:

  • ValueError: bad direction value

Returns:

  • np.ndarray: flipped image.

function Pad

Pad(img, dst_w, dst_h, pad_val=0, padding_mode='constant')

Padding a image to target size.

Args:

  • src (np.ndarray): Source image.

  • dst_w (int): target width.

  • dst_h (int): target height.

  • pad_val (int, optional): padding value to be filled in. Defaults to 0. User should set the value to some value meant to be ignored.

  • padding_mode (str): Type of padding. Should be: constant, edge,

  • reflect or symmetric. Default: constant.

    - constant: pads with a constant value, this value is specified

    with pad_val.

    - edge: pads with the last value at the edge of the image.

- reflect: pads with reflection of image without repeating the

    last value on the edge. For example, padding [1, 2, 3, 4] with 2 elements on both sides in reflect mode will result in [3, 2, 1, 2, 3, 4, 3, 2].

    - symmetric: pads with reflection of image repeating the last

    value on the edge. For example, padding [1, 2, 3, 4] with 2 elements on both sides in symmetric mode will result in [2, 1, 1, 2, 3, 4, 4, 3]

Returns:

  • np.ndarray: padded image.

function SizeDivisorMultiple

SizeDivisorMultiple(img=None, w=None, h=None, divisor)

Returns a smallest but larger shape to ensure each edge to be multiple to some number.

Args:

  • img (np.ndarray, optional): (H, W) or (H, W, 3), will extract image size if w or h not specified.

  • w (int, optional): original width

  • h (int, optional): original height

  • divisor (int, optional): the returned edge is a multiple of this value.

Returns:

  • dict: {"h": new_h, "w": new_w}