Box#

Bases: object

A multi-purpose box (ie. rectangle) representation .

Attributes

`area`	Return area of Box.
`height`	Return height of Box.
`size`
`width`	Return width of Box.

__init__(ymin, xmin, ymax, xmax)[source]#

Construct a bounding box.

Unless otherwise stated, the convention is that these coordinates are in pixel coordinates and represent boxes that lie within a RasterSource.

Parameters

ymin – minimum y value (y is row)
xmin – minimum x value (x is column)
ymax – maximum y value
xmax – maximum x value

Methods

`__init__`(ymin, xmin, ymax, xmax)	Construct a bounding box.
`buffer`(buffer_sz, max_extent)	Return new Box whose sides are buffered by buffer_sz.
`center_crop`(edge_offset_y, edge_offset_x)	Return Box whose sides are eroded by the given offsets.
`copy`()
`erode`(erosion_sz)	Return new Box whose sides are eroded by erosion_sz.
`filter_by_aoi`(windows, aoi_polygons[, within])	Filters windows by a list of AOI polygons
`from_dict`(d)
`from_npbox`(npbox)	Return new Box based on npbox format.
`from_rasterio`(rio_window)
`from_shapely`(shape)
`geojson_coordinates`()	Return Box as GeoJSON coordinates.
`get_windows`(size, stride[, padding, ...])	Returns a list of boxes representing windows generated using a sliding window traversal with the specified size, stride, and padding.
`intersection`(other)	Return the intersection of this Box and the other.
`intersects`(other)
`make_random_box_container`(out_h, out_w)	Return a new rectangular Box that contains this Box.
`make_random_square`(size)	Return new randomly positioned square Box that lies inside this Box.
`make_random_square_container`(size)	Return a new square Box that contains this Box.
`make_square`(ymin, xmin, size)	Return new square Box.
`npbox_format`()	Return Box in npbox format used by TF Object Detection API.
`pad`(ymin, xmin, ymax, xmax)	Pad sides by the given amount.
`rasterio_format`()	Return Box in Rasterio format: ((ymin, ymax), (xmin, xmax)).
`reproject`(transform_fn)	Reprojects this box based on a transform function.
`shapely_format`()
`shift_origin`(extent)	Shift origin of window coords to (extent.xmin, extent.ymin).
`to_dict`()
`to_int`()
`to_npboxes`(boxes)	Return nx4 numpy array from list of Box.
`to_offsets`(container)	Convert coords to offsets from (container.xmin, container.ymin).
`to_points`()	Get (x, y) coords of each vertex as a 4x2 numpy array.
`to_rasterio`()	Convert to a Rasterio Window.
`to_shapely`()	Convert to shapely Polygon.
`to_slices`()	Convert to slices: ymin:ymax, xmin:xmax
`to_xywh`()
`to_xyxy`()
`translate`(dy, dx)	Translate window along y and x axes by the given distances.
`tuple_format`()
`within_aoi`(window, aoi_polygons)	Check if window is within a list of AOI polygons.

__init__(ymin, xmin, ymax, xmax)[source]#

Construct a bounding box.

Unless otherwise stated, the convention is that these coordinates are in pixel coordinates and represent boxes that lie within a RasterSource.

Parameters

ymin – minimum y value (y is row)
xmin – minimum x value (x is column)
ymax – maximum y value
xmax – maximum x value

buffer(buffer_sz: float, max_extent: Box) → Box[source]#

Return new Box whose sides are buffered by buffer_sz.

The resulting box is clipped so that the values of the corners are always greater than zero and less than the height and width of max_extent.

Parameters

buffer_sz (float) –
max_extent (Box) –

Return type

Box

center_crop(edge_offset_y: int, edge_offset_x: int) → Box[source]#

Return Box whose sides are eroded by the given offsets.

Box(0, 0, 10, 10).center_crop(2, 4) == Box(2, 4, 8, 6)

Parameters

edge_offset_y (int) –
edge_offset_x (int) –

Return type

Box

copy() → Box[source]#

Return type: Box

erode(erosion_sz) → Box[source]#

Return new Box whose sides are eroded by erosion_sz.

Return type: Box

static filter_by_aoi(windows: List[Box], aoi_polygons: List[Polygon], within: bool = True) → List[Box][source]#

Filters windows by a list of AOI polygons

Parameters

within (bool) – if True, windows are only kept if they lie fully within an AOI polygon. Otherwise, windows are kept if they intersect an AOI polygon.
windows (List[Box]) –
aoi_polygons (List[Polygon]) –

Return type

List[Box]

classmethod from_dict(d: dict) → Box[source]#

Parameters: d (dict) –
Return type: Box

static from_npbox(npbox)[source]#

Return new Box based on npbox format.

Parameters: npbox – Numpy array of form [ymin, xmin, ymax, xmax] with float type

classmethod from_rasterio(rio_window: Window) → Box[source]#

Parameters: rio_window (Window) –
Return type: Box

static from_shapely(shape)[source]#

geojson_coordinates() → List[Tuple[int, int]][source]#

Return Box as GeoJSON coordinates.

Return type: List[Tuple[int, int]]

get_windows(size: Union[PositiveInt, Tuple[PositiveInt, PositiveInt]], stride: Union[PositiveInt, Tuple[PositiveInt, PositiveInt]], padding: Optional[Union[ConstrainedIntValue, Tuple[ConstrainedIntValue, ConstrainedIntValue]]] = None, pad_direction: Literal['both', 'start', 'end'] = 'end') → List[Box][source]#

Returns a list of boxes representing windows generated using a sliding window traversal with the specified size, stride, and padding.

Each of size, stride, and padding can be either a positive int or a tuple (vertical-componet, horizontal-component) of positive ints.

Padding currently only applies to the right and bottom edges.

Parameters

size (Union[PosInt, Tuple[PosInt, PosInt]]) – Size (h, w) of the windows.
stride (Union[PosInt, Tuple[PosInt, PosInt]]) – Step size between windows. Can be 2-tuple (h_step, w_step) or positive int.
padding (Optional[Union[PosInt, Tuple[PosInt, PosInt]]], optional) – Optional padding to accomodate windows that overflow the extent. Can be 2-tuple (h_pad, w_pad) or non-negative int. If None, will be set to (size[0]//2, size[1]//2). Defaults to None.
pad_direction (Literal['both', 'start', 'end']) – If ‘end’, only pad ymax and xmax (bottom and right). If ‘start’, only pad ymin and xmin (top and left). If ‘both’, pad all sides. Has no effect if paddiong is zero. Defaults to ‘end’.

Returns

List of Box objects.

Return type

List[Box]

intersection(other: Box) → Box[source]#

Return the intersection of this Box and the other.

Parameters: other (Box) – The box to intersect with this one.
Returns: The intersection of this box and the other one.
Return type: Box

intersects(other: Box) → bool[source]#

Parameters: other (Box) –
Return type: bool

make_random_box_container(out_h: int, out_w: int) → Box[source]#

Return a new rectangular Box that contains this Box.

Parameters

out_h (int) – the height of the new Box
out_w (int) – the width of the new Box

Return type

Box

make_random_square(size: int) → Box[source]#

Return new randomly positioned square Box that lies inside this Box.

Parameters: size (int) – the height and width of the new Box
Return type: Box

make_random_square_container(size: int) → Box[source]#

Return a new square Box that contains this Box.

Parameters: size (int) – the width and height of the new Box
Return type: Box

static make_square(ymin, xmin, size) → Box[source]#

Return new square Box.

Return type: Box

npbox_format()[source]#

Return Box in npbox format used by TF Object Detection API.

Returns: Numpy array of form [ymin, xmin, ymax, xmax] with float type

pad(ymin: int, xmin: int, ymax: int, xmax: int) → Box[source]#

Pad sides by the given amount.

Parameters

ymin (int) –
xmin (int) –
ymax (int) –
xmax (int) –

Return type

Box

rasterio_format() → Tuple[Tuple[int, int], Tuple[int, int]][source]#

Return Box in Rasterio format: ((ymin, ymax), (xmin, xmax)).

Return type: Tuple[Tuple[int, int], Tuple[int, int]]

reproject(transform_fn: Callable) → Box[source]#

Reprojects this box based on a transform function.

Parameters

tuple (transform_fn - A function that takes in a) – and reprojects that point to the target coordinate reference system.
transform_fn (Callable) –

Return type

Box

shapely_format() → Tuple[int, int, int, int][source]#