robokudo.types.cv¶

Computer vision types for Robokudo.

This module provides types for computer vision operations including:

2D and 3D point representations
Rectangle and region of interest definitions
3D bounding box specifications

The types support integration with: * OpenCV for image processing * Open3D for point cloud handling * Transform system for poses

Classes¶

`Point2D`	2D point representation.
`Points3D`	3D point cloud container.
`Rect`	2D rectangle representation.
`ImageROI`	Image region of interest.
`BoundingBox3D`	3D oriented bounding box.
`KeyPoint`	A key point in an image.

Module Contents¶

class robokudo.types.cv.Point2D¶

Bases: robokudo.types.core.Type

2D point representation.

Represents a point in 2D image coordinates.

x: int = 0¶: X coordinate

y: int = 0¶: Y coordinate

class robokudo.types.cv.Points3D¶

Bases: robokudo.types.core.Type

3D point cloud container.

Wraps an Open3D point cloud for 3D point operations.

points: open3d.geometry.PointCloud = None¶: The actual Open3D point cloud object

class robokudo.types.cv.Rect¶

Bases: robokudo.types.core.Type

2D rectangle representation.

Defines a rectangle by its top-left corner position and dimensions.

pos: Point2D¶: Top-left corner position

width: int = 0¶: Rectangle width in pixels

height: int = 0¶: Rectangle height in pixels

get_corner_points() → typing_extensions.Tuple[int, int, int, int]¶: Get the rectangle as a tuple of (x1, y1, x2, y2).

as_tuple() → typing_extensions.Tuple[int, int, int, int]¶: Get the rectangle as a tuple of (x, y, w, h).

class robokudo.types.cv.ImageROI¶

Bases: robokudo.types.core.Type

Image region of interest.

Defines a region of interest in an image using: * Binary mask for arbitrary shapes * Rectangle for bounding region

mask: numpy.typing.NDArray = None¶: Binary opencv mask image

roi: Rect¶: Rectangular region of interest

class robokudo.types.cv.BoundingBox3D¶

Bases: robokudo.types.core.Type

3D oriented bounding box.

Represents a 3D box with: * Dimensions along each axis * 6-DOF pose defining orientation and position

x_length: float = 0.0¶: Box length along x-axis

y_length: float = 0.0¶: Box length along y-axis

z_length: float = 0.0¶: Box length along z-axis

pose: robokudo.types.tf.Pose¶: Box pose in 3D space

class robokudo.types.cv.KeyPoint¶

Bases: Point2D

A key point in an image.

size: float¶: Diameter of the keypoint.

angle: float¶: Orientation of the keypoint.

response: float¶: Strength of the keypoint.

octave: int¶: Pyramid octave the keypoint was detected in.

class_id: int¶: Object class ID.

classmethod from_cv(keypoint: cv2.KeyPoint) → KeyPoint¶

Create a keypoint from an OpenCV keypoint.

Returns:: The created keypoint.

to_cv() → cv2.KeyPoint¶

Convert the RoboKudo keypoint to an OpenCV keypoint.

Returns:: The cv2 equivalent of the keypoint.