dgs.utils.validation.validate_bboxes

dgs.utils.validation.validate_bboxes(bboxes: torchvision.tv_tensors.BoundingBoxes, length: int | None = None, dims: int | None = 2, box_format: torchvision.tv_tensors.BoundingBoxFormat | None = None) → torchvision.tv_tensors.BoundingBoxes[source]

Given a torchvision tensor of bounding boxes, validate them and return them as a torchvision-tensor of bounding-boxes.

Parameters:

  • bboxes – tv_tensor.BoundingBoxes object with an arbitrary shape, most likely [B x 4].

  • length – The number of items or batch-size the tensor should have. Default None does not validate the length.

  • dims – Number of dimensions bboxes should have. Use None to not force any number of dimensions. Defaults to two dimensions with the bounding box dimensions as [B x 4].

  • box_format – If present, validates whether the tv_tensors.BoundingBoxFormat matches the one of bboxes. Default None, and therefore no validation of the format.

Returns:

Bounding boxes as tv_tensor.BoundingBoxes object with exactly dims dimensions.

Raises:

  • TypeError – If the bboxes input is not a Tensor.

  • ValueError – If the bboxes have the wrong shape or the bboxes have the wrong format.