from typing import Optional, List
import numpy as np
import cv2 as cv
from dataclasses import dataclass
from stardust.components.annotations import Box3D, Point3D
from stardust.components.camera.camera import Camera
from stardust.components.attachment.attachment import AttachmentType, UriInfo
ALL_IMAGE = (".jpg", ".jpeg", ".png", ".webp")
[docs]
class Image(UriInfo):
def __init__(
self,
uri: str,
width: Optional[int] = None,
height: Optional[int] = None,
name: Optional[str] = None,
camera_param: Optional[Camera] = None
):
"""
Picture object
Args:
width:
the picture width
height:
the picture height
name:
the picture name
camera_param:
class Camera
Returns:
Image instance
Examples:
.. code-block:: python
image = Image('', camera_param=cam)
"""
super(Image, self).__init__(uri)
self.type = AttachmentType.Image.value
self.width = width
self.height = height
self.name = name
self.camera_param = camera_param
@staticmethod
def _visual(**kwargs):
img = cv.imread(kwargs['file_path'])
cv.rectangle(img, kwargs['min_point'], kwargs['max_point'], (0, 255, 0), thickness=3, lineType=cv.LINE_AA)
cv.imshow('Image', img)
cv.waitKey(0)
@staticmethod
def _visual2(*args):
box1, box2, file_path = args
img = cv.imread(file_path)
for i in range(4):
cv.line(img, tuple(box1[i]), tuple(box1[(i + 1) % 4]), (0, 255, 0), 3)
cv.line(img, tuple(box2[i]), tuple(box2[(i + 1) % 4]), (0, 255, 0), 3)
Image.draw_dashed_line(img, tuple(box1[i]), tuple(box2[i]), (0, 255, 0), 3, line_type=cv.LINE_AA, shift=0)
cv.imshow('Image', img)
cv.waitKey(0)
cv.destroyAllWindows()
[docs]
@staticmethod
def draw_dashed_line(image, start_point, end_point, color, thickness, line_type, shift):
line_length = int(np.sqrt((end_point[0] - start_point[0]) ** 2 + (end_point[1] - start_point[1]) ** 2))
x_step = (end_point[0] - start_point[0]) / line_length
y_step = (end_point[1] - start_point[1]) / line_length
for i in range(0, line_length, 10): # 步长为10，可以根据需要调整
x1 = int(start_point[0] + i * x_step)
y1 = int(start_point[1] + i * y_step)
x2 = int(start_point[0] + (i + 5) * x_step)
y2 = int(start_point[1] + (i + 5) * y_step)
cv.line(image, (x1, y1), (x2, y2), color, thickness, lineType=line_type, shift=shift)
[docs]
@staticmethod
def get_points(max_point, min_point):
x1, y1 = np.array(min_point, int)
x2, y2 = np.array(max_point, int)
rectangle_coordinates = [(x1, y1), (x2, y1), (x2, y2), (x1, y2)]
return rectangle_coordinates
[docs]
def visual_2d_box(self, box3d: Box3D, file_path: Optional[str]):
vertices = box3d.vertices
vertices_matrix = np.c_[np.array([vertices[n] for n in [0, 3, 6, 1]]), np.array([1, 1, 1, 1])]
pixel_points = np.array([self.camera_param.get_pixel_coordinate(_[:3]) for _ in
np.dot(vertices_matrix, self.camera_param.cam2lidar_rt.I.T).tolist()], dtype=np.int32)
max_point = max(pixel_points, key=lambda x: sum(x))
min_point = min(pixel_points, key=lambda x: sum(x))
file_path = file_path if file_path else self.download_url()
self._visual(file_path=file_path, max_point=max_point, min_point=min_point)
[docs]
def visual_3d_box(self, box3d: Box3D, file_path: Optional[str]):
vertices = box3d.vertices
vertices_matrix = np.c_[np.array(vertices), np.array([1] * 8)]
pixel_points = np.array([self.camera_param.get_pixel_coordinate(_[:3]) for _ in
np.dot(vertices_matrix, self.camera_param.cam2lidar_rt.I.T).tolist()])
box1 = np.array([pixel_points[_] for _ in [0, 1, 3, 6]])
sum_of_coordinates1 = np.sum(box1, axis=1)
max_coordinate1 = box1[np.argmax(sum_of_coordinates1)]
min_coordinate1 = box1[np.argmin(sum_of_coordinates1)]
box1 = self.get_points(max_coordinate1, min_coordinate1)
box2 = np.array([pixel_points[_] for _ in [2, 4, 5, 7]])
sum_of_coordinates2 = np.sum(box2, axis=1)
max_coordinate2 = box2[np.argmax(sum_of_coordinates2)]
min_coordinate2 = box2[np.argmin(sum_of_coordinates2)]
box2 = self.get_points(max_coordinate2, min_coordinate2)
self._visual2(box1, box2, file_path)
[docs]
@staticmethod
def gen_image(image_obj):
"""
generate the Image obj
Args:
image_obj:
rosetta image data
Returns:
a Image instance
"""
_ = Image(uri=image_obj['url'],
width=image_obj.get('width'),
height=image_obj.get('height'),
name=image_obj['name'])
if "camera" in image_obj:
_.camera_param = Camera.gen_camera(image_obj['camera'])
return _
[docs]
@dataclass
class ImageSet:
"""
Args:
image_set: Image list
"""
type = AttachmentType.ImageSet.value
image_set: List[Image]
if __name__ == '__main__':
box3d = Box3D(
center=Point3D(19.617147983589785, -0.3571670183003431, 0.2083393465328241),
size=Point3D(2.0032044225562577, 4.793909959547556, 1.5418855556278792),
rotation=[0, 0, -1.5731558280731583]
)
cam = Camera(
'PinHole',
heading=(0.4351436321314605, -0.4677553765588058, 0.5492505771794342, -0.5386824023335351),
position=(1.5601294, -0.37986112, 1.3333596),
intrinsic=(1846.008599353862, 1857.930995475229, 1948.09353722228, 1078.768188621484),
radial=(0, 0, 0, 0),
tangential=(0, 0)
)
image = Image('', camera_param=cam)
# https://rosettalab.top/task/970/224049822964001714/viewOnly
# image.visual_2d_box(box3d, '/Users/stardust/Desktop/1681532060061245.jpg')
image.visual_3d_box(box3d, '/Users/mac/Downloads/jc93_suz_Ew_0b_sunny_m_0_1701929360184.jpg')