tlc.core.builtins.types.bb_crop_interface¶

Module Contents¶

Classes¶

Class

Description

BBCropInterface

Interface for creating bounding box crops.

API¶

class BBCropInterface¶

Interface for creating bounding box crops.

static crop(
image_path: str | Url | Image,
bb_dict: dict[str, float | int],
bb_schema: Schema,
image_height: int = 0,
image_width: int = 0,
x_max_offset: float = 0.0,
y_max_offset: float = 0.0,
y_scale_range: tuple[float, float] = (1.0, 1.0),
x_scale_range: tuple[float, float] = (1.0, 1.0),
) → Image¶

Crops an image according to a bounding box and returns the cropped image.

The parameters x_max_offset, y_max_offset, y_scale_range, and x_scale_range are used to introduce random variations in the crop, which can be useful for data augmentation during training.

Parameters:

  • image_path – Path to the image to crop.

  • bb_dict – Dictionary containing bounding box coordinates under the keys X0, Y0, X1, Y1.

  • bb_schema – Schema for the bounding box.

  • image_height – Height of the original image (only necessary if box is in relative coordinates).

  • image_width – Width of the original image (only necessary if box is in relative coordinates).

  • x_max_offset – Maximum random relative offset of the crop in x direction (both left and right).

  • y_max_offset – Maximum random relative offset of the crop in y direction (both up and down).

  • y_scale_range – Range of random relative scaling of the crop in y direction. The first value is the minimum scaling factor, the second value is the maximum scaling factor.

  • x_scale_range – Range of random relative scaling of the crop in x direction. The first value is the minimum scaling factor, the second value is the maximum scaling factor.

Returns:

Cropped image.