ConnectedComponentPostprocess

Note

This page is a reference documentation. It only explains the class signature, and not how to use it. Please refer to the Habitat Guide and Python API guide (v2.0) for usage.

class ConnectedComponentPostprocess(*, min_component_size: int = 30, connectivity: int = 1, reassign_method: str = 'neighbor_vote', max_iterations: int = 3)[source]

Bases: object

Clean tiny connected components inside an ROI label map.

Parameters:

  • min_component_size – Minimum voxels required to keep a component.

  • connectivity – Neighborhood connectivity in {1, 2, 3}.

  • reassign_method – Strategy name reserved for future variants; only "neighbor_vote" (nearest-seed refill) is implemented.

  • max_iterations – Cleanup iteration cap retained for YAML parity; the current kernel performs a single remove+refill pass.

__init__(*, min_component_size: int = 30, connectivity: int = 1, reassign_method: str = 'neighbor_vote', max_iterations: int = 3) → None[source]
apply_to_label_array(label_map: ndarray, roi_mask: ndarray) → ndarray[source]

Clean a raw label array inside roi_mask.

Parameters:

  • label_map – 3D integer label map (0 = background).

  • roi_mask – 3D boolean ROI mask.

Returns:

Cleaned int32 label map.

apply_to_habitat_map(habitat_map: HabitatMap) → HabitatMap[source]

Return a cleaned copy of a habitat label map.

ROI is label_array > 0. habitat_ids stay the model’s assignable set (unchanged), matching the contract that ids are model capability rather than presence in one subject.

Parameters:

habitat_map – Habitat labels for one subject.

Returns:

New HabitatMap with cleaned labels and derived provenance.

apply_to_supervoxelization(units: Supervoxelization, field: VoxelFeatureField) → Supervoxelization[source]

Clean a supervoxel partition and re-aggregate feature means.

Features must be recomputed after label reassignment so the feature matrix stays row-aligned with the surviving supervoxel ids.

Parameters:

  • units – Supervoxel partition to clean.

  • field – Per-voxel features used to rebuild region means.

Returns:

New Supervoxelization with cleaned labels and means.

Raises:

HABITAPIError – If field and units disagree on subject or geometry shape.