Source code for imgaug.augmentables.bbs

"""Classes representing bounding boxes."""
from __future__ import print_function, division, absolute_import

import copy

import numpy as np
import skimage.draw
import skimage.measure

from .. import imgaug as ia
from .base import IAugmentable
from .utils import (normalize_shape, project_coords,
                    _remove_out_of_image_fraction_,
                    _normalize_shift_args)


# TODO functions: square(), to_aspect_ratio(), contains_point()
[docs]class BoundingBox(object): """Class representing bounding boxes. Each bounding box is parameterized by its top left and bottom right corners. Both are given as x and y-coordinates. The corners are intended to lie inside the bounding box area. As a result, a bounding box that lies completely inside the image but has maximum extensions would have coordinates ``(0.0, 0.0)`` and ``(W - epsilon, H - epsilon)``. Note that coordinates are saved internally as floats. Parameters ---------- x1 : number X-coordinate of the top left of the bounding box. y1 : number Y-coordinate of the top left of the bounding box. x2 : number X-coordinate of the bottom right of the bounding box. y2 : number Y-coordinate of the bottom right of the bounding box. label : None or str, optional Label of the bounding box, e.g. a string representing the class. """ def __init__(self, x1, y1, x2, y2, label=None): """Create a new BoundingBox instance.""" if x1 > x2: x2, x1 = x1, x2 if y1 > y2: y2, y1 = y1, y2 self.x1 = x1 self.y1 = y1 self.x2 = x2 self.y2 = y2 self.label = label @property def coords(self): """Get the top-left and bottom-right coordinates as one array. Added in 0.4.0. Returns ------- ndarray A ``(N, 2)`` numpy array with ``N=2`` containing the top-left and bottom-right coordinates. """ arr = np.empty((2, 2), dtype=np.float32) arr[0, :] = (self.x1, self.y1) arr[1, :] = (self.x2, self.y2) return arr @property def x1_int(self): """Get the x-coordinate of the top left corner as an integer. Returns ------- int X-coordinate of the top left corner, rounded to the closest integer. """ # use numpy's round to have consistent behaviour between python # versions return int(np.round(self.x1)) @property def y1_int(self): """Get the y-coordinate of the top left corner as an integer. Returns ------- int Y-coordinate of the top left corner, rounded to the closest integer. """ # use numpy's round to have consistent behaviour between python # versions return int(np.round(self.y1)) @property def x2_int(self): """Get the x-coordinate of the bottom left corner as an integer. Returns ------- int X-coordinate of the bottom left corner, rounded to the closest integer. """ # use numpy's round to have consistent behaviour between python # versions return int(np.round(self.x2)) @property def y2_int(self): """Get the y-coordinate of the bottom left corner as an integer. Returns ------- int Y-coordinate of the bottom left corner, rounded to the closest integer. """ # use numpy's round to have consistent behaviour between python # versions return int(np.round(self.y2)) @property def height(self): """Estimate the height of the bounding box. Returns ------- number Height of the bounding box. """ return self.y2 - self.y1 @property def width(self): """Estimate the width of the bounding box. Returns ------- number Width of the bounding box. """ return self.x2 - self.x1 @property def center_x(self): """Estimate the x-coordinate of the center point of the bounding box. Returns ------- number X-coordinate of the center point of the bounding box. """ return self.x1 + self.width/2 @property def center_y(self): """Estimate the y-coordinate of the center point of the bounding box. Returns ------- number Y-coordinate of the center point of the bounding box. """ return self.y1 + self.height/2 @property def area(self): """Estimate the area of the bounding box. Returns ------- number Area of the bounding box, i.e. ``height * width``. """ return self.height * self.width # TODO add test for tuple of number
[docs] def contains(self, other): """Estimate whether the bounding box contains a given point. Parameters ---------- other : tuple of number or imgaug.augmentables.kps.Keypoint Point to check for. Returns ------- bool ``True`` if the point is contained in the bounding box, ``False`` otherwise. """ if isinstance(other, tuple): x, y = other else: x, y = other.x, other.y return self.x1 <= x <= self.x2 and self.y1 <= y <= self.y2
[docs] def project_(self, from_shape, to_shape): """Project the bounding box onto a differently shaped image in-place. E.g. if the bounding box is on its original image at ``x1=(10 of 100 pixels)`` and ``y1=(20 of 100 pixels)`` and is projected onto a new image with size ``(width=200, height=200)``, its new position will be ``(x1=20, y1=40)``. (Analogous for ``x2``/``y2``.) This is intended for cases where the original image is resized. It cannot be used for more complex changes (e.g. padding, cropping). Added in 0.4.0. Parameters ---------- from_shape : tuple of int or ndarray Shape of the original image. (Before resize.) to_shape : tuple of int or ndarray Shape of the new image. (After resize.) Returns ------- imgaug.augmentables.bbs.BoundingBox ``BoundingBox`` instance with new coordinates. The object may have been modified in-place. """ (self.x1, self.y1), (self.x2, self.y2) = project_coords( [(self.x1, self.y1), (self.x2, self.y2)], from_shape, to_shape) return self
# TODO add tests for ndarray inputs
[docs] def project(self, from_shape, to_shape): """Project the bounding box onto a differently shaped image. E.g. if the bounding box is on its original image at ``x1=(10 of 100 pixels)`` and ``y1=(20 of 100 pixels)`` and is projected onto a new image with size ``(width=200, height=200)``, its new position will be ``(x1=20, y1=40)``. (Analogous for ``x2``/``y2``.) This is intended for cases where the original image is resized. It cannot be used for more complex changes (e.g. padding, cropping). Parameters ---------- from_shape : tuple of int or ndarray Shape of the original image. (Before resize.) to_shape : tuple of int or ndarray Shape of the new image. (After resize.) Returns ------- imgaug.augmentables.bbs.BoundingBox ``BoundingBox`` instance with new coordinates. """ return self.deepcopy().project_(from_shape, to_shape)
[docs] def extend_(self, all_sides=0, top=0, right=0, bottom=0, left=0): """Extend the size of the bounding box along its sides in-place. Added in 0.4.0. Parameters ---------- all_sides : number, optional Value by which to extend the bounding box size along all sides. top : number, optional Value by which to extend the bounding box size along its top side. right : number, optional Value by which to extend the bounding box size along its right side. bottom : number, optional Value by which to extend the bounding box size along its bottom side. left : number, optional Value by which to extend the bounding box size along its left side. Returns ------- imgaug.BoundingBox Extended bounding box. The object may have been modified in-place. """ self.x1 = self.x1 - all_sides - left self.x2 = self.x2 + all_sides + right self.y1 = self.y1 - all_sides - top self.y2 = self.y2 + all_sides + bottom return self
[docs] def extend(self, all_sides=0, top=0, right=0, bottom=0, left=0): """Extend the size of the bounding box along its sides. Parameters ---------- all_sides : number, optional Value by which to extend the bounding box size along all sides. top : number, optional Value by which to extend the bounding box size along its top side. right : number, optional Value by which to extend the bounding box size along its right side. bottom : number, optional Value by which to extend the bounding box size along its bottom side. left : number, optional Value by which to extend the bounding box size along its left side. Returns ------- imgaug.BoundingBox Extended bounding box. """ return self.deepcopy().extend_(all_sides, top, right, bottom, left)
[docs] def intersection(self, other, default=None): """Compute the intersection BB between this BB and another BB. Note that in extreme cases, the intersection can be a single point. In that case the intersection bounding box exists and it will be returned, but it will have a height and width of zero. Parameters ---------- other : imgaug.augmentables.bbs.BoundingBox Other bounding box with which to generate the intersection. default : any, optional Default value to return if there is no intersection. Returns ------- imgaug.augmentables.bbs.BoundingBox or any Intersection bounding box of the two bounding boxes if there is an intersection. If there is no intersection, the default value will be returned, which can by anything. """ x1_i = max(self.x1, other.x1) y1_i = max(self.y1, other.y1) x2_i = min(self.x2, other.x2) y2_i = min(self.y2, other.y2) if x1_i > x2_i or y1_i > y2_i: return default return BoundingBox(x1=x1_i, y1=y1_i, x2=x2_i, y2=y2_i)
[docs] def union(self, other): """Compute the union BB between this BB and another BB. This is equivalent to drawing a bounding box around all corner points of both bounding boxes. Parameters ---------- other : imgaug.augmentables.bbs.BoundingBox Other bounding box with which to generate the union. Returns ------- imgaug.augmentables.bbs.BoundingBox Union bounding box of the two bounding boxes. """ return BoundingBox( x1=min(self.x1, other.x1), y1=min(self.y1, other.y1), x2=max(self.x2, other.x2), y2=max(self.y2, other.y2), )
[docs] def iou(self, other): """Compute the IoU between this bounding box and another one. IoU is the intersection over union, defined as:: ``area(intersection(A, B)) / area(union(A, B))`` ``= area(intersection(A, B)) / (area(A) + area(B) - area(intersection(A, B)))`` Parameters ---------- other : imgaug.augmentables.bbs.BoundingBox Other bounding box with which to compare. Returns ------- float IoU between the two bounding boxes. """ inters = self.intersection(other) if inters is None: return 0.0 area_union = self.area + other.area - inters.area return inters.area / area_union if area_union > 0 else 0.0
[docs] def compute_out_of_image_area(self, image): """Compute the area of the BB that is outside of the image plane. Added in 0.4.0. Parameters ---------- image : (H,W,...) ndarray or tuple of int Image dimensions to use. If an ``ndarray``, its shape will be used. If a ``tuple``, it is assumed to represent the image shape and must contain at least two integers. Returns ------- float Total area of the bounding box that is outside of the image plane. Can be ``0.0``. """ shape = normalize_shape(image) height, width = shape[0:2] bb_image = BoundingBox(x1=0, y1=0, x2=width, y2=height) inter = self.intersection(bb_image, default=None) area = self.area return area if inter is None else area - inter.area
[docs] def compute_out_of_image_fraction(self, image): """Compute fraction of BB area outside of the image plane. This estimates ``f = A_ooi / A``, where ``A_ooi`` is the area of the bounding box that is outside of the image plane, while ``A`` is the total area of the bounding box. Added in 0.4.0. Parameters ---------- image : (H,W,...) ndarray or tuple of int Image dimensions to use. If an ``ndarray``, its shape will be used. If a ``tuple``, it is assumed to represent the image shape and must contain at least two integers. Returns ------- float Fraction of the bounding box area that is outside of the image plane. Returns ``0.0`` if the bounding box is fully inside of the image plane. If the bounding box has an area of zero, the result is ``1.0`` if its coordinates are outside of the image plane, otherwise ``0.0``. """ area = self.area if area == 0: shape = normalize_shape(image) height, width = shape[0:2] y1_outside = self.y1 < 0 or self.y1 >= height x1_outside = self.x1 < 0 or self.x1 >= width is_outside = (y1_outside or x1_outside) return 1.0 if is_outside else 0.0 return self.compute_out_of_image_area(image) / area
[docs] def is_fully_within_image(self, image): """Estimate whether the bounding box is fully inside the image area. Parameters ---------- image : (H,W,...) ndarray or tuple of int Image dimensions to use. If an ``ndarray``, its shape will be used. If a ``tuple``, it is assumed to represent the image shape and must contain at least two integers. Returns ------- bool ``True`` if the bounding box is fully inside the image area. ``False`` otherwise. """ shape = normalize_shape(image) height, width = shape[0:2] return ( self.x1 >= 0 and self.x2 < width and self.y1 >= 0 and self.y2 < height)
[docs] def is_partly_within_image(self, image): """Estimate whether the BB is at least partially inside the image area. Parameters ---------- image : (H,W,...) ndarray or tuple of int Image dimensions to use. If an ``ndarray``, its shape will be used. If a ``tuple``, it is assumed to represent the image shape and must contain at least two integers. Returns ------- bool ``True`` if the bounding box is at least partially inside the image area. ``False`` otherwise. """ shape = normalize_shape(image) height, width = shape[0:2] eps = np.finfo(np.float32).eps img_bb = BoundingBox(x1=0, x2=width-eps, y1=0, y2=height-eps) return self.intersection(img_bb) is not None
[docs] def is_out_of_image(self, image, fully=True, partly=False): """Estimate whether the BB is partially/fully outside of the image area. Parameters ---------- image : (H,W,...) ndarray or tuple of int Image dimensions to use. If an ``ndarray``, its shape will be used. If a ``tuple``, it is assumed to represent the image shape and must contain at least two integers. fully : bool, optional Whether to return ``True`` if the bounding box is fully outside of the image area. partly : bool, optional Whether to return ``True`` if the bounding box is at least partially outside fo the image area. Returns ------- bool ``True`` if the bounding box is partially/fully outside of the image area, depending on defined parameters. ``False`` otherwise. """ if self.is_fully_within_image(image): return False if self.is_partly_within_image(image): return partly return fully
[docs] @ia.deprecated(alt_func="BoundingBox.clip_out_of_image()", comment="clip_out_of_image() has the exactly same " "interface.") def cut_out_of_image(self, *args, **kwargs): """Clip off all parts of the BB box that are outside of the image.""" return self.clip_out_of_image(*args, **kwargs)
[docs] def clip_out_of_image_(self, image): """Clip off parts of the BB box that are outside of the image in-place. Added in 0.4.0. Parameters ---------- image : (H,W,...) ndarray or tuple of int Image dimensions to use for the clipping of the bounding box. If an ``ndarray``, its shape will be used. If a ``tuple``, it is assumed to represent the image shape and must contain at least two integers. Returns ------- imgaug.augmentables.bbs.BoundingBox Bounding box, clipped to fall within the image dimensions. The object may have been modified in-place. """ shape = normalize_shape(image) height, width = shape[0:2] assert height > 0, ( "Expected image with height>0, got shape %s." % (image.shape,)) assert width > 0, ( "Expected image with width>0, got shape %s." % (image.shape,)) eps = np.finfo(np.float32).eps self.x1 = np.clip(self.x1, 0, width - eps) self.x2 = np.clip(self.x2, 0, width - eps) self.y1 = np.clip(self.y1, 0, height - eps) self.y2 = np.clip(self.y2, 0, height - eps) return self
[docs] def clip_out_of_image(self, image): """Clip off all parts of the BB box that are outside of the image. Parameters ---------- image : (H,W,...) ndarray or tuple of int Image dimensions to use for the clipping of the bounding box. If an ``ndarray``, its shape will be used. If a ``tuple``, it is assumed to represent the image shape and must contain at least two integers. Returns ------- imgaug.augmentables.bbs.BoundingBox Bounding box, clipped to fall within the image dimensions. """ return self.deepcopy().clip_out_of_image_(image)
[docs] def shift_(self, x=0, y=0): """Move this bounding box along the x/y-axis in-place. The origin ``(0, 0)`` is at the top left of the image. Added in 0.4.0. Parameters ---------- x : number, optional Value to be added to all x-coordinates. Positive values shift towards the right images. y : number, optional Value to be added to all y-coordinates. Positive values shift towards the bottom images. Returns ------- imgaug.augmentables.bbs.BoundingBox Shifted bounding box. The object may have been modified in-place. """ self.x1 += x self.x2 += x self.y1 += y self.y2 += y return self
[docs] def shift(self, x=0, y=0, top=None, right=None, bottom=None, left=None): """Move this bounding box along the x/y-axis. The origin ``(0, 0)`` is at the top left of the image. Parameters ---------- x : number, optional Value to be added to all x-coordinates. Positive values shift towards the right images. y : number, optional Value to be added to all y-coordinates. Positive values shift towards the bottom images. top : None or int, optional Deprecated since 0.4.0. Amount of pixels by which to shift this object *from* the top (towards the bottom). right : None or int, optional Deprecated since 0.4.0. Amount of pixels by which to shift this object *from* the right (towards the left). bottom : None or int, optional Deprecated since 0.4.0. Amount of pixels by which to shift this object *from* the bottom (towards the top). left : None or int, optional Deprecated since 0.4.0. Amount of pixels by which to shift this object *from* the left (towards the right). Returns ------- imgaug.augmentables.bbs.BoundingBox Shifted bounding box. """ # pylint: disable=redefined-outer-name x, y = _normalize_shift_args( x, y, top=top, right=right, bottom=bottom, left=left) return self.deepcopy().shift_(x, y)
[docs] def draw_label_on_image(self, image, color=(0, 255, 0), color_text=None, color_bg=None, alpha=1.0, size=1, size_text=20, height=30, copy=True, raise_if_out_of_image=False): """Draw a box showing the BB's label. The box is placed right above the BB's rectangle. Added in 0.4.0. Parameters ---------- image : (H,W,C) ndarray The image onto which to draw the label. Currently expected to be ``uint8``. color : None or iterable of int, optional The color to use, corresponding to the channel layout of the image. Usually RGB. Text and background colors will be derived from this. color_text : None or iterable of int, optional The text color to use. If ``None``, derived from `color_bg`. color_bg : None or iterable of int, optional The background color of the label box. If ``None``, derived from `color`. alpha : float, optional The transparency of the drawn bounding box, where ``1.0`` denotes no transparency and ``0.0`` is invisible. size : int, optional The thickness of the bounding box in pixels. If the value is larger than ``1``, then additional pixels will be added around the bounding box (i.e. extension towards the outside). size_text : int, optional Font size to use. height : int, optional Height of the label box in pixels. copy : bool, optional Whether to copy the input image or change it in-place. raise_if_out_of_image : bool, optional Whether to raise an error if the bounding box is fully outside of the image. If set to ``False``, no error will be raised and only the parts inside the image will be drawn. Returns ------- (H,W,C) ndarray(uint8) Image with bounding box drawn on it. """ # pylint: disable=redefined-outer-name drawer = _LabelOnImageDrawer( color=color, color_text=color_text, color_bg=color_bg, size=size, alpha=alpha, raise_if_out_of_image=raise_if_out_of_image, height=height, size_text=size_text) if copy: return drawer.draw_on_image(image, self) return drawer.draw_on_image_(image, self)
[docs] def draw_box_on_image(self, image, color=(0, 255, 0), alpha=1.0, size=1, copy=True, raise_if_out_of_image=False, thickness=None): """Draw the rectangle of the bounding box on an image. This method does not draw the label. Added in 0.4.0. Parameters ---------- image : (H,W,C) ndarray The image onto which to draw the bounding box rectangle. Currently expected to be ``uint8``. color : iterable of int, optional The color to use, corresponding to the channel layout of the image. Usually RGB. alpha : float, optional The transparency of the drawn bounding box, where ``1.0`` denotes no transparency and ``0.0`` is invisible. size : int, optional The thickness of the bounding box in pixels. If the value is larger than ``1``, then additional pixels will be added around the bounding box (i.e. extension towards the outside). copy : bool, optional Whether to copy the input image or change it in-place. raise_if_out_of_image : bool, optional Whether to raise an error if the bounding box is fully outside of the image. If set to ``False``, no error will be raised and only the parts inside the image will be drawn. thickness : None or int, optional Deprecated. Returns ------- (H,W,C) ndarray(uint8) Image with bounding box drawn on it. """ # pylint: disable=invalid-name, redefined-outer-name if thickness is not None: ia.warn_deprecated( "Usage of argument 'thickness' in BoundingBox.draw_on_image() " "is deprecated. The argument was renamed to 'size'.") size = thickness if raise_if_out_of_image and self.is_out_of_image(image): raise Exception( "Cannot draw bounding box x1=%.8f, y1=%.8f, x2=%.8f, y2=%.8f " "on image with shape %s." % ( self.x1, self.y1, self.x2, self.y2, image.shape)) result = np.copy(image) if copy else image if isinstance(color, (tuple, list)): color = np.uint8(color) for i in range(size): y1, y2, x1, x2 = self.y1_int, self.y2_int, self.x1_int, self.x2_int # When y values get into the range (H-0.5, H), the *_int functions # round them to H. That is technically sensible, but in the case # of drawing means that the border lies just barely outside of # the image, making the border disappear, even though the BB is # fully inside the image. Here we correct for that because of # beauty reasons. Same is the case for x coordinates. if self.is_fully_within_image(image): y1 = np.clip(y1, 0, image.shape[0]-1) y2 = np.clip(y2, 0, image.shape[0]-1) x1 = np.clip(x1, 0, image.shape[1]-1) x2 = np.clip(x2, 0, image.shape[1]-1) y = [y1-i, y1-i, y2+i, y2+i] x = [x1-i, x2+i, x2+i, x1-i] rr, cc = skimage.draw.polygon_perimeter(y, x, shape=result.shape) if alpha >= 0.99: result[rr, cc, :] = color else: if ia.is_float_array(result): # TODO use blend_alpha here result[rr, cc, :] = ( (1 - alpha) * result[rr, cc, :] + alpha * color) result = np.clip(result, 0, 255) else: input_dtype = result.dtype result = result.astype(np.float32) result[rr, cc, :] = ( (1 - alpha) * result[rr, cc, :] + alpha * color) result = np.clip(result, 0, 255).astype(input_dtype) return result
# TODO add explicit test for zero-sized BBs (worked when tested by hand)
[docs] def draw_on_image(self, image, color=(0, 255, 0), alpha=1.0, size=1, copy=True, raise_if_out_of_image=False, thickness=None): """Draw the bounding box on an image. This will automatically also draw the label, unless it is ``None``. To only draw the box rectangle use :func:`~imgaug.augmentables.bbs.BoundingBox.draw_box_on_image`. To draw the label even if it is ``None`` or to configure e.g. its color, use :func:`~imgaug.augmentables.bbs.BoundingBox.draw_label_on_image`. Parameters ---------- image : (H,W,C) ndarray The image onto which to draw the bounding box. Currently expected to be ``uint8``. color : iterable of int, optional The color to use, corresponding to the channel layout of the image. Usually RGB. alpha : float, optional The transparency of the drawn bounding box, where ``1.0`` denotes no transparency and ``0.0`` is invisible. size : int, optional The thickness of the bounding box in pixels. If the value is larger than ``1``, then additional pixels will be added around the bounding box (i.e. extension towards the outside). copy : bool, optional Whether to copy the input image or change it in-place. raise_if_out_of_image : bool, optional Whether to raise an error if the bounding box is fully outside of the image. If set to ``False``, no error will be raised and only the parts inside the image will be drawn. thickness : None or int, optional Deprecated. Returns ------- (H,W,C) ndarray(uint8) Image with bounding box drawn on it. """ # pylint: disable=redefined-outer-name image_drawn = self.draw_box_on_image( image, color=color, alpha=alpha, size=size, copy=copy, raise_if_out_of_image=raise_if_out_of_image, thickness=thickness ) if self.label is not None: image_drawn = self.draw_label_on_image( image_drawn, color=color, alpha=alpha, size=size if thickness is None else thickness, copy=False, raise_if_out_of_image=raise_if_out_of_image ) return image_drawn
# TODO add tests for pad and pad_max
[docs] def extract_from_image(self, image, pad=True, pad_max=None, prevent_zero_size=True): """Extract the image pixels within the bounding box. This function will zero-pad the image if the bounding box is partially/fully outside of the image. Parameters ---------- image : (H,W) ndarray or (H,W,C) ndarray The image from which to extract the pixels within the bounding box. pad : bool, optional Whether to zero-pad the image if the object is partially/fully outside of it. pad_max : None or int, optional The maximum number of pixels that may be zero-paded on any side, i.e. if this has value ``N`` the total maximum of added pixels is ``4*N``. This option exists to prevent extremely large images as a result of single points being moved very far away during augmentation. prevent_zero_size : bool, optional Whether to prevent the height or width of the extracted image from becoming zero. If this is set to ``True`` and the height or width of the bounding box is below ``1``, the height/width will be increased to ``1``. This can be useful to prevent problems, e.g. with image saving or plotting. If it is set to ``False``, images will be returned as ``(H', W')`` or ``(H', W', 3)`` with ``H`` or ``W`` potentially being 0. Returns ------- (H',W') ndarray or (H',W',C) ndarray Pixels within the bounding box. Zero-padded if the bounding box is partially/fully outside of the image. If `prevent_zero_size` is activated, it is guarantueed that ``H'>0`` and ``W'>0``, otherwise only ``H'>=0`` and ``W'>=0``. """ # pylint: disable=no-else-return, too-many-statements height, width = image.shape[0], image.shape[1] x1, x2, y1, y2 = self.x1_int, self.x2_int, self.y1_int, self.y2_int # When y values get into the range (H-0.5, H), the *_int functions # round them to H. That is technically sensible, but in the case of # extraction leads to a black border, which is both ugly and # unexpected after calling cut_out_of_image(). Here we correct for # that because of beauty reasons. Same is the case for x coordinates. fully_within = self.is_fully_within_image(image) if fully_within: y1, y2 = np.clip([y1, y2], 0, height-1) x1, x2 = np.clip([x1, x2], 0, width-1) # TODO add test if prevent_zero_size: if abs(x2 - x1) < 1: x2 = x1 + 1 if abs(y2 - y1) < 1: y2 = y1 + 1 if pad: # if the bb is outside of the image area, the following pads the # image first with black pixels until the bb is inside the image # and only then extracts the image area # TODO probably more efficient to initialize an array of zeros # and copy only the portions of the bb into that array that # are natively inside the image area from ..augmenters import size as iasize pad_top = 0 pad_right = 0 pad_bottom = 0 pad_left = 0 if x1 < 0: pad_left = abs(x1) x2 = x2 + pad_left width = width + pad_left x1 = 0 if y1 < 0: pad_top = abs(y1) y2 = y2 + pad_top height = height + pad_top y1 = 0 if x2 >= width: pad_right = x2 - width if y2 >= height: pad_bottom = y2 - height paddings = [pad_top, pad_right, pad_bottom, pad_left] any_padded = any([val > 0 for val in paddings]) if any_padded: if pad_max is None: pad_max = max(paddings) image = iasize.pad( image, top=min(pad_top, pad_max), right=min(pad_right, pad_max), bottom=min(pad_bottom, pad_max), left=min(pad_left, pad_max) ) return image[y1:y2, x1:x2] else: within_image = ( (0, 0, 0, 0) <= (x1, y1, x2, y2) < (width, height, width, height) ) out_height, out_width = (y2 - y1), (x2 - x1) nonzero_height = (out_height > 0) nonzero_width = (out_width > 0) if within_image and nonzero_height and nonzero_width: return image[y1:y2, x1:x2] if prevent_zero_size: out_height = 1 out_width = 1 else: out_height = 0 out_width = 0 if image.ndim == 2: return np.zeros((out_height, out_width), dtype=image.dtype) return np.zeros((out_height, out_width, image.shape[-1]), dtype=image.dtype)
# TODO also add to_heatmap # TODO add this to BoundingBoxesOnImage # TODO add label to keypoints?
[docs] def to_keypoints(self): """Convert the BB's corners to keypoints (clockwise, from top left). Returns ------- list of imgaug.augmentables.kps.Keypoint Corners of the bounding box as keypoints. """ # TODO get rid of this deferred import from imgaug.augmentables.kps import Keypoint return [ Keypoint(x=self.x1, y=self.y1), Keypoint(x=self.x2, y=self.y1), Keypoint(x=self.x2, y=self.y2), Keypoint(x=self.x1, y=self.y2) ]
[docs] def to_polygon(self): """Convert this bounding box to a polygon covering the same area. Added in 0.4.0. Returns ------- imgaug.augmentables.polys.Polygon The bounding box converted to a polygon. """ # TODO get rid of this deferred import from imgaug.augmentables.polys import Polygon return Polygon([ (self.x1, self.y1), (self.x2, self.y1), (self.x2, self.y2), (self.x1, self.y2) ], label=self.label)
# TODO also introduce similar area_almost_equals()
[docs] def coords_almost_equals(self, other, max_distance=1e-4): """Estimate if this and another BB have almost identical coordinates. Added in 0.4.0. Parameters ---------- other : imgaug.augmentables.bbs.BoundingBox or iterable The other bounding box with which to compare this one. If this is an ``iterable``, it is assumed to represent the top-left and bottom-right coordinates of that bounding box, given as e.g. an ``(2,2)`` ndarray or an ``(4,)`` ndarray or as a similar list. max_distance : number, optional The maximum euclidean distance between a corner on one bounding box and the closest corner on the other bounding box. If the distance is exceeded for any such pair, the two BBs are not viewed as equal. Returns ------- bool Whether the two bounding boxes have almost identical corner coordinates. """ if isinstance(other, BoundingBox): coords_b = other.coords.flat elif ia.is_np_array(other): # we use flat here in case other is (N,2) instead of (4,) coords_b = other.flat elif ia.is_iterable(other): coords_b = list(ia.flatten(other)) else: raise ValueError( "Expected 'other' to be an iterable containing two " "(x,y)-coordinate pairs or a BoundingBox. " "Got type %s." % (type(other),)) coords_a = self.coords return np.allclose(coords_a.flat, coords_b, atol=max_distance, rtol=0)
[docs] def almost_equals(self, other, max_distance=1e-4): """Compare this and another BB's label and coordinates. This is the same as :func:`~imgaug.augmentables.bbs.BoundingBox.coords_almost_equals` but additionally compares the labels. Added in 0.4.0. Parameters ---------- other : imgaug.augmentables.bbs.BoundingBox or iterable The other object to compare against. Expected to be a ``BoundingBox``. max_distance : number, optional See :func:`~imgaug.augmentables.bbs.BoundingBox.coords_almost_equals`. Returns ------- bool ``True`` if the coordinates are almost equal and additionally the labels are equal. Otherwise ``False``. """ if self.label != other.label: return False return self.coords_almost_equals(other, max_distance=max_distance)
[docs] @classmethod def from_point_soup(cls, xy): """Convert a ``(2P,) or (P,2) ndarray`` to a BB instance. This is the inverse of :func:`~imgaug.BoundingBoxesOnImage.to_xyxy_array`. Added in 0.4.0. Parameters ---------- xy : (2P,) ndarray or (P, 2) array or iterable of number or iterable of iterable of number Array containing ``P`` points in xy-form denoting a soup of points around which to place a bounding box. The array should usually be of dtype ``float32``. Returns ------- imgaug.augmentables.bbs.BoundingBox Bounding box around the points. """ # pylint: disable=unsubscriptable-object xy = np.array(xy, dtype=np.float32) assert len(xy) > 0, ( "Expected to get at least one point to place a bounding box " "around, got shape %s." % (xy.shape,)) assert xy.ndim == 1 or (xy.ndim == 2 and xy.shape[-1] == 2), ( "Expected input array of shape (P,) or (P, 2), " "got shape %s." % (xy.shape,)) if xy.ndim == 1: xy = xy.reshape((-1, 2)) x1, y1 = np.min(xy, axis=0) x2, y2 = np.max(xy, axis=0) return cls(x1=x1, y1=y1, x2=x2, y2=y2)
[docs] def copy(self, x1=None, y1=None, x2=None, y2=None, label=None): """Create a shallow copy of this BoundingBox instance. Parameters ---------- x1 : None or number If not ``None``, then the ``x1`` coordinate of the copied object will be set to this value. y1 : None or number If not ``None``, then the ``y1`` coordinate of the copied object will be set to this value. x2 : None or number If not ``None``, then the ``x2`` coordinate of the copied object will be set to this value. y2 : None or number If not ``None``, then the ``y2`` coordinate of the copied object will be set to this value. label : None or string If not ``None``, then the ``label`` of the copied object will be set to this value. Returns ------- imgaug.augmentables.bbs.BoundingBox Shallow copy. """ return BoundingBox( x1=self.x1 if x1 is None else x1, x2=self.x2 if x2 is None else x2, y1=self.y1 if y1 is None else y1, y2=self.y2 if y2 is None else y2, label=copy.deepcopy(self.label) if label is None else label )
[docs] def deepcopy(self, x1=None, y1=None, x2=None, y2=None, label=None): """ Create a deep copy of the BoundingBox object. Parameters ---------- x1 : None or number If not ``None``, then the ``x1`` coordinate of the copied object will be set to this value. y1 : None or number If not ``None``, then the ``y1`` coordinate of the copied object will be set to this value. x2 : None or number If not ``None``, then the ``x2`` coordinate of the copied object will be set to this value. y2 : None or number If not ``None``, then the ``y2`` coordinate of the copied object will be set to this value. label : None or string If not ``None``, then the ``label`` of the copied object will be set to this value. Returns ------- imgaug.augmentables.bbs.BoundingBox Deep copy. """ # TODO write specific copy routine with deepcopy for label and remove # the deepcopy from copy() return self.copy(x1=x1, y1=y1, x2=x2, y2=y2, label=label)
def __getitem__(self, indices): """Get the coordinate(s) with given indices. Added in 0.4.0. Returns ------- ndarray xy-coordinate(s) as ``ndarray``. """ return self.coords[indices] def __iter__(self): """Iterate over the coordinates of this instance. Added in 0.4.0. Yields ------ ndarray An ``(2,)`` ``ndarray`` denoting an xy-coordinate pair. """ return iter(self.coords) def __repr__(self): return self.__str__() def __str__(self): return "BoundingBox(x1=%.4f, y1=%.4f, x2=%.4f, y2=%.4f, label=%s)" % ( self.x1, self.y1, self.x2, self.y2, self.label)
[docs]class BoundingBoxesOnImage(IAugmentable): """Container for the list of all bounding boxes on a single image. Parameters ---------- bounding_boxes : list of imgaug.augmentables.bbs.BoundingBox List of bounding boxes on the image. shape : tuple of int or ndarray The shape of the image on which the objects are placed. Either an image with shape ``(H,W,[C])`` or a ``tuple`` denoting such an image shape. Examples -------- >>> import numpy as np >>> from imgaug.augmentables.bbs import BoundingBox, BoundingBoxesOnImage >>> >>> image = np.zeros((100, 100)) >>> bbs = [ >>> BoundingBox(x1=10, y1=20, x2=20, y2=30), >>> BoundingBox(x1=25, y1=50, x2=30, y2=70) >>> ] >>> bbs_oi = BoundingBoxesOnImage(bbs, shape=image.shape) """ def __init__(self, bounding_boxes, shape): self.bounding_boxes = bounding_boxes self.shape = normalize_shape(shape) @property def items(self): """Get the bounding boxes in this container. Added in 0.4.0. Returns ------- list of BoundingBox Bounding boxes within this container. """ return self.bounding_boxes @items.setter def items(self, value): """Set the bounding boxes in this container. Added in 0.4.0. Parameters ---------- value : list of BoundingBox Bounding boxes within this container. """ self.bounding_boxes = value # TODO remove this? here it is image height, but in BoundingBox it is # bounding box height @property def height(self): """Get the height of the image on which the bounding boxes fall. Returns ------- int Image height. """ return self.shape[0] # TODO remove this? here it is image width, but in BoundingBox it is # bounding box width @property def width(self): """Get the width of the image on which the bounding boxes fall. Returns ------- int Image width. """ return self.shape[1] @property def empty(self): """Determine whether this instance contains zero bounding boxes. Returns ------- bool True if this object contains zero bounding boxes. """ return len(self.bounding_boxes) == 0
[docs] def on_(self, image): """Project BBs from one image (shape) to a another one in-place. Added in 0.4.0. Parameters ---------- image : ndarray or tuple of int New image onto which the bounding boxes are to be projected. May also simply be that new image's shape tuple. Returns ------- imgaug.augmentables.bbs.BoundingBoxesOnImage Object containing the same bounding boxes after projection to the new image shape. The object and its items may have been modified in-place. """ # pylint: disable=invalid-name on_shape = normalize_shape(image) if on_shape[0:2] == self.shape[0:2]: self.shape = on_shape # channels may differ return self for i, item in enumerate(self.items): self.bounding_boxes[i] = item.project_(self.shape, on_shape) self.shape = on_shape return self
[docs] def on(self, image): """Project bounding boxes from one image (shape) to a another one. Parameters ---------- image : ndarray or tuple of int New image onto which the bounding boxes are to be projected. May also simply be that new image's shape tuple. Returns ------- imgaug.augmentables.bbs.BoundingBoxesOnImage Object containing the same bounding boxes after projection to the new image shape. """ # pylint: disable=invalid-name return self.deepcopy().on_(image)
[docs] @classmethod def from_xyxy_array(cls, xyxy, shape): """Convert an ``(N, 4) or (N, 2, 2) ndarray`` to a BBsOI instance. This is the inverse of :func:`~imgaug.BoundingBoxesOnImage.to_xyxy_array`. Parameters ---------- xyxy : (N, 4) ndarray or (N, 2, 2) array Array containing the corner coordinates of ``N`` bounding boxes. Each bounding box is represented by its top-left and bottom-right coordinates. The array should usually be of dtype ``float32``. shape : tuple of int Shape of the image on which the bounding boxes are placed. Should usually be ``(H, W, C)`` or ``(H, W)``. Returns ------- imgaug.augmentables.bbs.BoundingBoxesOnImage Object containing a list of :class:`BoundingBox` instances derived from the provided corner coordinates. """ # pylint: disable=unsubscriptable-object xyxy = np.array(xyxy, dtype=np.float32) # note that np.array([]) is (0,), not (0, 2) if xyxy.shape[0] == 0: return BoundingBoxesOnImage([], shape) assert ( (xyxy.ndim == 2 and xyxy.shape[-1] == 4) or (xyxy.ndim == 3 and xyxy.shape[1:3] == (2, 2))), ( "Expected input array of shape (N, 4) or (N, 2, 2), " "got shape %s." % (xyxy.shape,)) xyxy = xyxy.reshape((-1, 2, 2)) boxes = [BoundingBox.from_point_soup(row) for row in xyxy] return cls(boxes, shape)
[docs] @classmethod def from_point_soups(cls, xy, shape): """Convert an ``(N, 2P) or (N, P, 2) ndarray`` to a BBsOI instance. Added in 0.4.0. Parameters ---------- xy : (N, 2P) ndarray or (N, P, 2) array or iterable of iterable of number or iterable of iterable of iterable of number Array containing the corner coordinates of ``N`` bounding boxes. Each bounding box is represented by a soup of ``P`` points. If ``(N, P)`` then the second axis is expected to be in xy-form (e.g. ``x1``, ``y1``, ``x2``, ``y2``, ...). The final bounding box coordinates will be derived using ``min`` and ``max`` operations on the xy-values. The array should usually be of dtype ``float32``. shape : tuple of int Shape of the image on which the bounding boxes are placed. Should usually be ``(H, W, C)`` or ``(H, W)``. Returns ------- imgaug.augmentables.bbs.BoundingBoxesOnImage Object containing a list of :class:`BoundingBox` instances derived from the provided point soups. """ xy = np.array(xy, dtype=np.float32) # from_xy_array() already checks the ndim/shape, so we don't have to # do it here boxes = [BoundingBox.from_point_soup(row) for row in xy] return cls(boxes, shape)
[docs] def to_xyxy_array(self, dtype=np.float32): """Convert the ``BoundingBoxesOnImage`` object to an ``(N,4) ndarray``. This is the inverse of :func:`~imgaug.BoundingBoxesOnImage.from_xyxy_array`. Parameters ---------- dtype : numpy.dtype, optional Desired output datatype of the ndarray. Returns ------- ndarray ``(N,4) ndarray``, where ``N`` denotes the number of bounding boxes and ``4`` denotes the top-left and bottom-right bounding box corner coordinates in form ``(x1, y1, x2, y2)``. """ xyxy_array = np.zeros((len(self.bounding_boxes), 4), dtype=np.float32) for i, box in enumerate(self.bounding_boxes): xyxy_array[i] = [box.x1, box.y1, box.x2, box.y2] return xyxy_array.astype(dtype)
[docs] def to_xy_array(self): """Convert the ``BoundingBoxesOnImage`` object to an ``(N,2) ndarray``. Added in 0.4.0. Returns ------- ndarray ``(2*B,2) ndarray`` of xy-coordinates, where ``B`` denotes the number of bounding boxes. """ return self.to_xyxy_array().reshape((-1, 2))
[docs] def fill_from_xyxy_array_(self, xyxy): """Modify the BB coordinates of this instance in-place. .. note:: This currently expects exactly one entry in `xyxy` per bounding in this instance. (I.e. two corner coordinates per instance.) Otherwise, an ``AssertionError`` will be raised. .. note:: This method will automatically flip x-coordinates if ``x1>x2`` for a bounding box. (Analogous for y-coordinates.) Added in 0.4.0. Parameters ---------- xyxy : (N, 4) ndarray or iterable of iterable of number Coordinates of ``N`` bounding boxes on an image, given as a ``(N,4)`` array of two corner xy-coordinates per bounding box. ``N`` must match the number of bounding boxes in this instance. Returns ------- BoundingBoxesOnImage This instance itself, with updated bounding box coordinates. Note that the instance was modified in-place. """ xyxy = np.array(xyxy, dtype=np.float32) # note that np.array([]) is (0,), not (0, 4) assert xyxy.shape[0] == 0 or (xyxy.ndim == 2 and xyxy.shape[-1] == 4), ( # pylint: disable=unsubscriptable-object "Expected input array to have shape (N,4), " "got shape %s." % (xyxy.shape,)) assert len(xyxy) == len(self.bounding_boxes), ( "Expected to receive an array with as many rows there are " "bounding boxes in this instance. Got %d rows, expected %d." % ( len(xyxy), len(self.bounding_boxes))) for bb, (x1, y1, x2, y2) in zip(self.bounding_boxes, xyxy): bb.x1 = min([x1, x2]) bb.y1 = min([y1, y2]) bb.x2 = max([x1, x2]) bb.y2 = max([y1, y2]) return self
[docs] def fill_from_xy_array_(self, xy): """Modify the BB coordinates of this instance in-place. See :func:`~imgaug.augmentables.bbs.BoundingBoxesOnImage.fill_from_xyxy_array_`. Added in 0.4.0. Parameters ---------- xy : (2*B, 2) ndarray or iterable of iterable of number Coordinates of ``B`` bounding boxes on an image, given as a ``(2*B,2)`` array of two corner xy-coordinates per bounding box. ``B`` must match the number of bounding boxes in this instance. Returns ------- BoundingBoxesOnImage This instance itself, with updated bounding box coordinates. Note that the instance was modified in-place. """ xy = np.array(xy, dtype=np.float32) return self.fill_from_xyxy_array_(xy.reshape((-1, 4)))
[docs] def draw_on_image(self, image, color=(0, 255, 0), alpha=1.0, size=1, copy=True, raise_if_out_of_image=False, thickness=None): """Draw all bounding boxes onto a given image. Parameters ---------- image : (H,W,3) ndarray The image onto which to draw the bounding boxes. This image should usually have the same shape as set in ``BoundingBoxesOnImage.shape``. color : int or list of int or tuple of int or (3,) ndarray, optional The RGB color of all bounding boxes. If a single ``int`` ``C``, then that is equivalent to ``(C,C,C)``. alpha : float, optional Alpha/transparency of the bounding box. size : int, optional Thickness in pixels. copy : bool, optional Whether to copy the image before drawing the bounding boxes. raise_if_out_of_image : bool, optional Whether to raise an exception if any bounding box is outside of the image. thickness : None or int, optional Deprecated. Returns ------- (H,W,3) ndarray Image with drawn bounding boxes. """ # pylint: disable=redefined-outer-name image = np.copy(image) if copy else image for bb in self.bounding_boxes: image = bb.draw_on_image( image, color=color, alpha=alpha, size=size, copy=False, raise_if_out_of_image=raise_if_out_of_image, thickness=thickness ) return image
[docs] def remove_out_of_image_(self, fully=True, partly=False): """Remove in-place all BBs that are fully/partially outside of the image. Added in 0.4.0. Parameters ---------- fully : bool, optional Whether to remove bounding boxes that are fully outside of the image. partly : bool, optional Whether to remove bounding boxes that are partially outside of the image. Returns ------- imgaug.augmentables.bbs.BoundingBoxesOnImage Reduced set of bounding boxes, with those that were fully/partially outside of the image being removed. The object and its items may have been modified in-place. """ self.bounding_boxes = [ bb for bb in self.bounding_boxes if not bb.is_out_of_image(self.shape, fully=fully, partly=partly)] return self
[docs] def remove_out_of_image(self, fully=True, partly=False): """Remove all BBs that are fully/partially outside of the image. Parameters ---------- fully : bool, optional Whether to remove bounding boxes that are fully outside of the image. partly : bool, optional Whether to remove bounding boxes that are partially outside of the image. Returns ------- imgaug.augmentables.bbs.BoundingBoxesOnImage Reduced set of bounding boxes, with those that were fully/partially outside of the image being removed. """ return self.copy().remove_out_of_image_(fully=fully, partly=partly)
[docs] def remove_out_of_image_fraction_(self, fraction): """Remove in-place all BBs with an OOI fraction of at least `fraction`. 'OOI' is the abbreviation for 'out of image'. Added in 0.4.0. Parameters ---------- fraction : number Minimum out of image fraction that a bounding box has to have in order to be removed. A fraction of ``1.0`` removes only bounding boxes that are ``100%`` outside of the image. A fraction of ``0.0`` removes all bounding boxes. Returns ------- imgaug.augmentables.bbs.BoundingBoxesOnImage Reduced set of bounding boxes, with those that had an out of image fraction greater or equal the given one removed. The object and its items may have been modified in-place. """ return _remove_out_of_image_fraction_(self, fraction)
[docs] def remove_out_of_image_fraction(self, fraction): """Remove all BBs with an out of image fraction of at least `fraction`. Added in 0.4.0. Parameters ---------- fraction : number Minimum out of image fraction that a bounding box has to have in order to be removed. A fraction of ``1.0`` removes only bounding boxes that are ``100%`` outside of the image. A fraction of ``0.0`` removes all bounding boxes. Returns ------- imgaug.augmentables.bbs.BoundingBoxesOnImage Reduced set of bounding boxes, with those that had an out of image fraction greater or equal the given one removed. """ return self.copy().remove_out_of_image_fraction_(fraction)
[docs] @ia.deprecated(alt_func="BoundingBoxesOnImage.clip_out_of_image()", comment="clip_out_of_image() has the exactly same " "interface.") def cut_out_of_image(self): """Clip off all parts from all BBs that are outside of the image.""" return self.clip_out_of_image()
[docs] def clip_out_of_image_(self): """ Clip off in-place all parts from all BBs that are outside of the image. Added in 0.4.0. Returns ------- imgaug.augmentables.bbs.BoundingBoxesOnImage Bounding boxes, clipped to fall within the image dimensions. The object and its items may have been modified in-place. """ # remove bbs that are not at least partially inside the image plane self.bounding_boxes = [bb for bb in self.bounding_boxes if bb.is_partly_within_image(self.shape)] for i, bb in enumerate(self.bounding_boxes): self.bounding_boxes[i] = bb.clip_out_of_image(self.shape) return self
[docs] def clip_out_of_image(self): """Clip off all parts from all BBs that are outside of the image. Returns ------- imgaug.augmentables.bbs.BoundingBoxesOnImage Bounding boxes, clipped to fall within the image dimensions. """ return self.deepcopy().clip_out_of_image_()
[docs] def shift_(self, x=0, y=0): """Move all BBs along the x/y-axis in-place. The origin ``(0, 0)`` is at the top left of the image. Added in 0.4.0. Parameters ---------- x : number, optional Value to be added to all x-coordinates. Positive values shift towards the right images. y : number, optional Value to be added to all y-coordinates. Positive values shift towards the bottom images. Returns ------- imgaug.augmentables.bbs.BoundingBoxesOnImage Shifted bounding boxes. The object and its items may have been modified in-place. """ for i, bb in enumerate(self.bounding_boxes): self.bounding_boxes[i] = bb.shift_(x=x, y=y) return self
[docs] def shift(self, x=0, y=0, top=None, right=None, bottom=None, left=None): """Move all BBs along the x/y-axis. The origin ``(0, 0)`` is at the top left of the image. Parameters ---------- x : number, optional Value to be added to all x-coordinates. Positive values shift towards the right images. y : number, optional Value to be added to all y-coordinates. Positive values shift towards the bottom images. top : None or int, optional Deprecated since 0.4.0. Amount of pixels by which to shift all objects *from* the top (towards the bottom). right : None or int, optional Deprecated since 0.4.0. Amount of pixels by which to shift all objects *from* the right (towads the left). bottom : None or int, optional Deprecated since 0.4.0. Amount of pixels by which to shift all objects *from* the bottom (towards the top). left : None or int, optional Deprecated since 0.4.0. Amount of pixels by which to shift all objects *from* the left (towards the right). Returns ------- imgaug.augmentables.bbs.BoundingBoxesOnImage Shifted bounding boxes. """ x, y = _normalize_shift_args( x, y, top=top, right=right, bottom=bottom, left=left) return self.deepcopy().shift_(x=x, y=y)
[docs] def to_keypoints_on_image(self): """Convert the bounding boxes to one ``KeypointsOnImage`` instance. Added in 0.4.0. Returns ------- imgaug.augmentables.kps.KeypointsOnImage A keypoints instance containing ``N*4`` coordinates for ``N`` bounding boxes. Order matches the order in ``bounding_boxes``. """ from .kps import KeypointsOnImage # This currently uses 4 points instead of 2 points as the method # is primarily used during augmentation and 4 points are overall # the better choice there. arr = np.zeros((len(self.bounding_boxes), 2*4), dtype=np.float32) for i, box in enumerate(self.bounding_boxes): arr[i] = [ box.x1, box.y1, box.x2, box.y1, box.x2, box.y2, box.x1, box.y2 ] return KeypointsOnImage.from_xy_array( arr.reshape((-1, 2)), shape=self.shape )
[docs] def invert_to_keypoints_on_image_(self, kpsoi): """Invert the output of ``to_keypoints_on_image()`` in-place. This function writes in-place into this ``BoundingBoxesOnImage`` instance. Added in 0.4.0. Parameters ---------- kpsoi : imgaug.augmentables.kps.KeypointsOnImages Keypoints to convert back to bounding boxes, i.e. the outputs of ``to_keypoints_on_image()``. Returns ------- BoundingBoxesOnImage Bounding boxes container with updated coordinates. Note that the instance is also updated in-place. """ assert len(kpsoi.keypoints) == len(self.bounding_boxes) * 4, ( "Expected %d coordinates, got %d." % ( len(self.bounding_boxes) * 2, len(kpsoi.keypoints))) for i, bb in enumerate(self.bounding_boxes): xx = [kpsoi.keypoints[4*i+0].x, kpsoi.keypoints[4*i+1].x, kpsoi.keypoints[4*i+2].x, kpsoi.keypoints[4*i+3].x] yy = [kpsoi.keypoints[4*i+0].y, kpsoi.keypoints[4*i+1].y, kpsoi.keypoints[4*i+2].y, kpsoi.keypoints[4*i+3].y] bb.x1 = min(xx) bb.y1 = min(yy) bb.x2 = max(xx) bb.y2 = max(yy) self.shape = kpsoi.shape return self
[docs] def to_polygons_on_image(self): """Convert the bounding boxes to one ``PolygonsOnImage`` instance. Added in 0.4.0. Returns ------- imgaug.augmentables.polys.PolygonsOnImage A ``PolygonsOnImage`` containing polygons. Each polygon covers the same area as the corresponding bounding box. """ from .polys import PolygonsOnImage polygons = [bb.to_polygon() for bb in self.bounding_boxes] return PolygonsOnImage(polygons, shape=self.shape)
[docs] def copy(self, bounding_boxes=None, shape=None): """Create a shallow copy of the ``BoundingBoxesOnImage`` instance. Parameters ---------- bounding_boxes : None or list of imgaug.augmntables.bbs.BoundingBox, optional List of bounding boxes on the image. If ``None``, the instance's bounding boxes will be copied. shape : tuple of int, optional The shape of the image on which the bounding boxes are placed. If ``None``, the instance's shape will be copied. Returns ------- imgaug.augmentables.bbs.BoundingBoxesOnImage Shallow copy. """ if bounding_boxes is None: bounding_boxes = self.bounding_boxes[:] if shape is None: # use tuple() here in case the shape was provided as a list shape = tuple(self.shape) return BoundingBoxesOnImage(bounding_boxes, shape)
[docs] def deepcopy(self, bounding_boxes=None, shape=None): """Create a deep copy of the ``BoundingBoxesOnImage`` object. Parameters ---------- bounding_boxes : None or list of imgaug.augmntables.bbs.BoundingBox, optional List of bounding boxes on the image. If ``None``, the instance's bounding boxes will be copied. shape : tuple of int, optional The shape of the image on which the bounding boxes are placed. If ``None``, the instance's shape will be copied. Returns ------- imgaug.augmentables.bbs.BoundingBoxesOnImage Deep copy. """ # Manual copy is far faster than deepcopy, so use manual copy here. if bounding_boxes is None: bounding_boxes = [bb.deepcopy() for bb in self.bounding_boxes] if shape is None: # use tuple() here in case the shape was provided as a list shape = tuple(self.shape) return BoundingBoxesOnImage(bounding_boxes, shape)
def __getitem__(self, indices): """Get the bounding box(es) with given indices. Added in 0.4.0. Returns ------- list of imgaug.augmentables.bbs.BoundingBoxes Bounding box(es) with given indices. """ return self.bounding_boxes[indices] def __iter__(self): """Iterate over the bounding boxes in this container. Added in 0.4.0. Yields ------ BoundingBox A bounding box in this container. The order is identical to the order in the bounding box list provided upon class initialization. """ return iter(self.bounding_boxes) def __len__(self): """Get the number of items in this instance. Added in 0.4.0. Returns ------- int Number of items in this instance. """ return len(self.items) def __repr__(self): return self.__str__() def __str__(self): return ( "BoundingBoxesOnImage(%s, shape=%s)" % (str(self.bounding_boxes), self.shape))
class _LabelOnImageDrawer(object): # size refers to the thickness of the BB # height is the height of the label rectangle, not the whole BB def __init__(self, color=(0, 255, 0), color_text=None, color_bg=None, size=1, alpha=1.0, raise_if_out_of_image=False, height=30, size_text=20): self.color = color self.color_text = color_text self.color_bg = color_bg self.size = size self.alpha = alpha self.raise_if_out_of_image = raise_if_out_of_image self.height = height self.size_text = size_text def draw_on_image_(self, image, bounding_box): # pylint: disable=invalid-name, redefined-outer-name if self.raise_if_out_of_image: self._do_raise_if_out_of_image(image, bounding_box) color_text, color_bg = self._preprocess_colors() x1, y1, x2, y2 = self._compute_bg_corner_coords(image, bounding_box) # cant draw anything if OOI if x2 <= x1 or y2 <= y1: return image # can currently only draw on images with shape (H,W,C), not (H,W) label_arr = self._draw_label_arr(bounding_box.label, y2 - y1, x2 - x1, image.shape[-1], image.dtype, color_text, color_bg, self.size_text) image = self._blend_label_arr_with_image_(image, label_arr, x1, y1, x2, y2) return image def draw_on_image(self, image, bounding_box): return self.draw_on_image_(np.copy(image), bounding_box) @classmethod def _do_raise_if_out_of_image(cls, image, bounding_box): if bounding_box.is_out_of_image(image): raise Exception( "Cannot draw bounding box x1=%.8f, y1=%.8f, x2=%.8f, y2=%.8f " "on image with shape %s." % ( bounding_box.x1, bounding_box.y1, bounding_box.x2, bounding_box.y2, image.shape)) def _preprocess_colors(self): color = np.uint8(self.color) if self.color is not None else None color_bg = self.color_bg if self.color_bg is not None: color_bg = np.uint8(color_bg) else: assert color is not None, ( "Expected `color` to be set when `color_bg` is not set, " "but it was None.") color_bg = color color_text = self.color_text if self.color_text is not None: color_text = np.uint8(color_text) else: # we follow the approach from https://stackoverflow.com/a/1855903 # here gray = (0.299 * color_bg[0] + 0.587 * color_bg[1] + 0.114 * color_bg[2]) color_text = np.full((3,), 0 if gray > 128 else 255, dtype=np.uint8) return color_text, color_bg def _compute_bg_corner_coords(self, image, bounding_box): bb = bounding_box offset = self.size height, width = image.shape[0:2] y1, x1, x2 = bb.y1_int, bb.x1_int, bb.x2_int # dont use bb.y2 here! we want the label to be above the BB y1 = y1 - 1 - self.height y2 = y1 + self.height x1 = x1 - offset + 1 x2 = x2 + offset y1, y2 = np.clip([y1, y2], 0, height-1) x1, x2 = np.clip([x1, x2], 0, width-1) return x1, y1, x2, y2 @classmethod def _draw_label_arr(cls, label, height, width, nb_channels, dtype, color_text, color_bg, size_text): label_arr = np.zeros((height, width, nb_channels), dtype=dtype) label_arr[...] = color_bg.reshape((1, 1, -1)) label_arr = ia.draw_text(label_arr, x=2, y=2, text=str(label), color=color_text, size=size_text) return label_arr def _blend_label_arr_with_image_(self, image, label_arr, x1, y1, x2, y2): alpha = self.alpha if alpha >= 0.99: image[y1:y2, x1:x2, :] = label_arr else: input_dtype = image.dtype foreground = label_arr.astype(np.float64) background = image[y1:y2, x1:x2, :].astype(np.float64) blend = (1 - alpha) * background + alpha * foreground blend = np.clip(blend, 0, 255).astype(input_dtype) image[y1:y2, x1:x2, :] = blend return image