robokudo.annotators.tsdf_annotator¶

Classes¶

`TSDFAnnotatorVisualizationModes`	Enumeration of supported visualization modes for TSDF volumes.
`TSDFAnnotator`	Annotator for per-object TSDF volume integration.

Module Contents¶

class robokudo.annotators.tsdf_annotator.TSDFAnnotatorVisualizationModes¶

Bases: enum.Enum

Enumeration of supported visualization modes for TSDF volumes.

NONE = 'none'¶

MESH = 'mesh'¶

POINTCLOUD = 'pointcloud'¶

VOXEL_POINTCLOUD = 'voxel_pointcloud'¶

class robokudo.annotators.tsdf_annotator.TSDFAnnotator(name: str = 'TSDFAnnotator', descriptor: TSDFAnnotator = Descriptor())¶

Bases: robokudo.annotators.core.ThreadedAnnotator

Annotator for per-object TSDF volume integration.

Note

The annotator requires a camera to world transform and object poses to integrate the TSDF volumes. To segment objects out of the image a mask is required too. Higher quality masks yield better results.

class Descriptor¶

Bases: robokudo.annotators.core.BaseAnnotator.Descriptor

Configuration descriptor for the TSDF annotator.

class Parameters¶

Bases: robokudo.annotators.core.BaseAnnotator.Descriptor.Parameters

Parameters for configuring the TSDF annotator.

voxel_length: float = 0.005¶: Voxel length for the TSDF volume of each object.

sdf_trunc: float = 0.04¶: SDF truncation distance for the TSDF volume of each object.

depth_scale: float = 1000.0¶: Depth scale for the TSDF volume integration.

depth_trunc: float = 3.0¶: Depth truncation distance for the TSDF volume integration.

geometry_visualization_mode: TSDFAnnotatorVisualizationModes¶: Whether to visualize the TSDF as a mesh or a point cloud.

parameters¶

translation_comparator¶: Translation comparator for comparing world positions between pipeline iterations.

tracked_objects: typing_extensions.List[typing_extensions.Dict[str, typing_extensions.Any]] = []¶: A list of tracked object data, including the bounding box and the TSDF volume.

find_tracked_object(pose: robokudo.types.annotation.PoseAnnotation, threshold: float = 0.75) → typing_extensions.Optional[typing_extensions.Dict[str, typing_extensions.Any]]¶

Check whether the given bounding box is similar to one that has been seen before.

Parameters:

pose – The pose to check
threshold – The similarity threshold between 0.0 and 1.0 for considering the bounding box similar.

Returns:

The most similar tracked object or none if no similar object was found.

compute() → py_trees.common.Status¶

Generate a TSDF volume for each object hypothesis or integrate into existing volumes.

Returns:: Whether the TSDF integration was successful or not.

get_visualization_image(image: numpy.typing.NDArray[numpy.uint8], oh_data: typing_extensions.List[typing_extensions.Tuple[robokudo.types.scene.ObjectHypothesis, robokudo.types.annotation.PoseAnnotation, typing_extensions.Tuple[int, int, int, int]]], masks: typing_extensions.List[numpy.typing.NDArray[numpy.uint8]]) → numpy.typing.NDArray[numpy.uint8]¶

Get an image containing the bounding boxes and masks of objects.

Parameters:

image – The image to add the visualizations to.
oh_data – The object hypotheses to visualize.
masks – The masks to visualize.

Returns:

The image with the visualizations.

get_visualization_geometries(volume: robokudo.types.cv.TSDFAnnotation) → typing_extensions.List[open3d.geometry.Geometry]¶

Get visualization geometries for the TSDF volume depending on the visualization mode.

Parameters:: volume – The TSDF volume annotation to visualize.
Returns:: List of visualization geometries for the given volume annotation.