BoxList#

class BoxList[source]#

Bases: object

__init__(boxes: Tensor, format: str = 'xyxy', **extras) → None[source]#

Representation of a list of bounding boxes and associated data.

Internally, boxes are always stored in the xyxy format.

Parameters:

  • boxes (Tensor) – tensor<n, 4>

  • format (str) – format of input boxes.

  • extras – dict with values that are tensors with first dimension corresponding to boxes first dimension

Return type:

None

Methods

__init__(boxes[, format])

Representation of a list of bounding boxes and associated data.

cat(box_lists)

clip_boxes(img_height, img_width)

convert_boxes(out_fmt)

copy()

equal(other)

get_field(name)

ind_filter(inds)

nms([iou_thresh])

pin_memory()

scale(yscale, xscale)

Scale box coords by the given scaling factors.

score_filter([score_thresh])

to(*args, **kwargs)

Recursively apply torch.Tensor.to() to Tensors.
__contains__(key: str) → bool[source]#
Parameters:

key (str) –

Return type:

bool

__init__(boxes: Tensor, format: str = 'xyxy', **extras) → None[source]#

Representation of a list of bounding boxes and associated data.

Internally, boxes are always stored in the xyxy format.

Parameters:

  • boxes (Tensor) – tensor<n, 4>

  • format (str) – format of input boxes.

  • extras – dict with values that are tensors with first dimension corresponding to boxes first dimension

Return type:

None

static cat(box_lists: Iterable[Self]) → Self[source]#
Parameters:

box_lists (Iterable[Self]) –

Return type:

Self

clip_boxes(img_height: int, img_width: int) → Self[source]#
Parameters:

  • img_height (int) –

  • img_width (int) –

Return type:

Self

convert_boxes(out_fmt: str) → Tensor[source]#
Parameters:

out_fmt (str) –

Return type:

Tensor

copy() → Self[source]#
Return type:

Self

equal(other: Self) → bool[source]#
Parameters:

other (Self) –

Return type:

bool

get_field(name: str) → Any[source]#
Parameters:

name (str) –

Return type:

Any

ind_filter(inds: Sequence[int]) → Self[source]#
Parameters:

inds (Sequence[int]) –

Return type:

Self

nms(iou_thresh: float = 0.5) → Tensor[source]#
Parameters:

iou_thresh (float) –

Return type:

Tensor

pin_memory() → Self[source]#
Return type:

Self

scale(yscale: float, xscale: float) → Self[source]#

Scale box coords by the given scaling factors.

Parameters:
Return type:

Self

score_filter(score_thresh: float = 0.25) → Self[source]#
Parameters:

score_thresh (float) –

Return type:

Self

to(*args, **kwargs) → Self[source]#

Recursively apply torch.Tensor.to() to Tensors.

Parameters:
Returns:

New BoxList with to’d Tensors.

Return type:

BoxList