1 vin.geometry.qh_queries

vin.geometry.qh_queries

Shared target-lattice and candidate-ray query geometry for QH readouts.

1.1 Classes

Name Description
QhSpatialQueryGeometry World-frame points and row support shared by spatial readouts.

1.1.1 QhSpatialQueryGeometry

vin.geometry.qh_queries.QhSpatialQueryGeometry(
    target_points,
    ray_points,
    candidate_centres,
    target_centre,
    materialized,
    candidate_pose_world,
    target_pose_world,
)

World-frame points and row support shared by spatial readouts.

1.1.1.1 Attributes

Name Description
target_points Tensor["B K_t 3", float] target-OBB lattice points in world metres.
ray_points Tensor["B S N K_r 3", float] candidate-to-target ray points.
candidate_centres Tensor["B S N 3", float] candidate optical centres in world metres.
target_centre Tensor["B 3", float] target centre in world metres.
materialized Tensor["B S N", bool] candidate and realized-state support.
candidate_pose_world PoseTW["B S N 12"] sanitized world-from-candidate poses.
target_pose_world PoseTW["B 12"] sanitized world-from-target pose.

1.2 Functions

Name Description
build_qh_spatial_queries Construct the existing target OBB lattice and candidate-target rays.

1.2.1 build_qh_spatial_queries

vin.geometry.qh_queries.build_qh_spatial_queries(
    actor,
    *,
    target_lattice,
    ray_fraction,
)

Construct the existing target OBB lattice and candidate-target rays.

This pure tensor/pose seam centralizes the query definition used by static EVL and learned-volume readouts. It deliberately does not construct a camera frustum because the candidate rows carry no camera calibration.

Back to top