dacapo.blockwise.empanada_function

Attributes

logger

default_parameters

model_configs

Functions

segment_function(input_array, block, **parameters)

Segment a 3D block using the empanada-napari library.

stack_inference(engine, volume, axis_name)

Perform inference on a single axis of a 3D volume.

orthoplane_inference(engine, volume)

Perform inference on the orthogonal planes of a 3D volume.

empanada_segmenter(image[, model_config, use_gpu, ...])

Segment a 3D volume using the empanada-napari library.

stack_postprocessing(trackers, model_config[, ...])

Relabels and filters each class defined in trackers. Yields a numpy

tracker_consensus(trackers, model_config[, ...])

Calculate the orthoplane consensus from trackers. Yields a numpy

Module Contents

dacapo.blockwise.empanada_function.logger
dacapo.blockwise.empanada_function.default_parameters
dacapo.blockwise.empanada_function.segment_function(input_array, block, **parameters)

Segment a 3D block using the empanada-napari library.

Parameters:

  • input_array (np.ndarray) – The 3D array to segment.

  • block (daisy.Block) – The block object.

  • **parameters – Parameters for the empanada-napari segmenter.

Returns:

The segmented 3D array.

Return type:

np.ndarray

Raises:

ImportError – If empanada-napari is not installed.

Examples

>>> import numpy as np
>>> from dask import array as da
>>> from dacapo.blockwise.empanada_function import segment_function
>>> input_array = np.random.rand(64, 64, 64)
>>> block = da.from_array(input_array, chunks=(32, 32, 32))
>>> segmented_array = segment_function(block, model_config="MitoNet_v1")

Note

The model_config parameter should be one of the following: - MitoNet_v1 - MitoNet_v2 - MitoNet_v3 - MitoNet_v4 - MitoNet_v5 - MitoNet_v6

Reference:

  • doi: 10.1016/j.cels.2022.12.006

dacapo.blockwise.empanada_function.model_configs
dacapo.blockwise.empanada_function.stack_inference(engine, volume, axis_name)

Perform inference on a single axis of a 3D volume.

Parameters:

  • engine (Engine3d) – The engine object.

  • volume (np.ndarray) – The 3D volume to segment.

  • axis_name (str) – The axis name to segment.

Returns:

The stack, axis name, and trackers dictionary.

Return type:

tuple

Raises:

ImportError – If empanada-napari is not installed.

Examples

>>> import numpy as np
>>> from empanada_napari.inference import Engine3d
>>> from dacapo.blockwise.empanada_function import stack_inference
>>> model_config = "MitoNet_v1"
>>> use_gpu = True
>>> use_quantized = False
>>> engine = Engine3d(model_config, use_gpu=use_gpu, use_quantized=use_quantized)
>>> volume = np.random.rand(64, 64, 64)
>>> axis_name = "xy"
>>> stack, axis_name, trackers_dict = stack_inference(engine, volume, axis_name)

Note

The axis_name parameter should be one of the following:

dacapo.blockwise.empanada_function.orthoplane_inference(engine, volume)

Perform inference on the orthogonal planes of a 3D volume.

Parameters:

  • engine (Engine3d) – The engine object.

  • volume (np.ndarray) – The 3D volume to segment.

Returns:

The trackers dictionary.

Return type:

dict

Raises:

ImportError – If empanada-napari is not installed.

Examples

>>> import numpy as np
>>> from empanada_napari.inference import Engine3d
>>> from dacapo.blockwise.empanada_function import orthoplane_inference
>>> model_config = "MitoNet_v1"
>>> use_gpu = True
>>> use_quantized = False
>>> engine = Engine3d(model_config, use_gpu=use_gpu, use_quantized=use_quantized)
>>> volume = np.random.rand(64, 64, 64)
>>> trackers_dict = orthoplane_inference(engine, volume)

Note

The model_config parameter should be one of the following: - MitoNet_v1 - MitoNet_v2 - MitoNet_v3 - MitoNet_v4 - MitoNet_v5 - MitoNet_v6

dacapo.blockwise.empanada_function.empanada_segmenter(image, model_config='MitoNet_v1', use_gpu=True, use_quantized=False, multigpu=False, downsampling=1, confidence_thr=0.5, center_confidence_thr=0.1, min_distance_object_centers=21, fine_boundaries=True, semantic_only=False, median_slices=11, min_size=10000, min_extent=50, maximum_objects_per_class=1000000, inference_plane='xy', orthoplane=True, return_panoptic=False, pixel_vote_thr=1, allow_one_view=False)

Segment a 3D volume using the empanada-napari library.

Parameters:

  • image (np.ndarray) – The 3D volume to segment.

  • model_config (str) – The model configuration to use.

  • use_gpu (bool) – Whether to use the GPU.

  • use_quantized (bool) – Whether to use quantized inference.

  • multigpu (bool) – Whether to use multiple GPUs.

  • downsampling (int) – The downsampling factor.

  • confidence_thr (float) – The confidence threshold.

  • center_confidence_thr (float) – The center confidence threshold.

  • min_distance_object_centers (int) – The minimum distance between object centers.

  • fine_boundaries (bool) – Whether to use fine boundaries.

  • semantic_only (bool) – Whether to use semantic segmentation only.

  • median_slices (int) – The number of median slices.

  • min_size (int) – The minimum size of objects.

  • min_extent (int) – The minimum extent.

  • maximum_objects_per_class (int) – The maximum number of objects per class.

  • inference_plane (str) – The inference plane.

  • orthoplane (bool) – Whether to use orthoplane inference.

  • return_panoptic (bool) – Whether to return the panoptic segmentation.

  • pixel_vote_thr (int) – The pixel vote threshold.

  • allow_one_view (bool) – Whether to allow one view.

Returns:

The volume, class name, and tracker.

Return type:

tuple

Raises:

ImportError – If empanada-napari is not installed.

Examples

>>> import numpy as np
>>> from empanada_napari.inference import Engine3d
>>> from dacapo.blockwise.empanada_function import empanada_segmenter
>>> image = np.random.rand(64, 64, 64)
>>> model_config = "MitoNet_v1"
>>> use_gpu = True
>>> use_quantized = False
>>> multigpu = False
>>> downsampling = 1
>>> confidence_thr = 0.5
>>> center_confidence_thr = 0.1
>>> min_distance_object_centers = 21
>>> fine_boundaries = True
>>> semantic_only = False
>>> median_slices = 11
>>> min_size = 10000
>>> min_extent = 50
>>> maximum_objects_per_class = 1000000
>>> inference_plane = "xy"
>>> orthoplane = True
>>> return_panoptic = False
>>> pixel_vote_thr = 1
>>> allow_one_view = False
>>> for vol, class_name, tracker in empanada_segmenter(
...     image,
...     model_config=model_config,
...     use_gpu=use_gpu,
...     use_quantized=use_quantized,
...     multigpu=multigpu,
...     downsampling=downsampling,
...     confidence_thr=confidence_thr,
...     center_confidence_thr=center_confidence_thr,
...     min_distance_object_centers=min_distance_object_centers,
...     fine_boundaries=fine_boundaries,
...     semantic_only=semantic_only,
...     median_slices=median_slices,
...     min_size=min_size,
...     min_extent=min_extent,
...     maximum_objects_per_class=maximum_objects_per_class,
...     inference_plane=inference_plane,
...     orthoplane=orthoplane,
...     return_panoptic=return_panoptic,
...     pixel_vote_thr=pixel_vote_thr,
...     allow_one_view=allow_one_view
... ):
...     print(vol.shape, class_name, tracker)

Note

The model_config parameter should be one of the following: - MitoNet_v1 - MitoNet_v2 - MitoNet_v3 - MitoNet_v4 - MitoNet_v5 - MitoNet_v6

Reference:

  • doi: 10.1016/j.cels.2022.12.006

dacapo.blockwise.empanada_function.stack_postprocessing(trackers, model_config, label_divisor=1000, min_size=200, min_extent=4, dtype=np.uint32)

Relabels and filters each class defined in trackers. Yields a numpy or zarr volume along with the name of the class that is segmented.

Parameters:

  • trackers (dict) – The trackers dictionary.

  • model_config (str) – The model configuration to use.

  • label_divisor (int) – The label divisor.

  • min_size (int) – The minimum size of objects.

  • min_extent (int) – The minimum extent of objects.

  • dtype (type) – The data type.

Returns:

The generator object.

Return type:

generator

Raises:

ImportError – If empanada-napari is not installed.

Examples

>>> for vol, class_name, tracker in stack_postprocessing(trackers, model_config):
...     print(vol.shape, class_name, tracker)

Note

The model_config parameter should be one of the following: - MitoNet_v1 - MitoNet_v2 - MitoNet_v3 - MitoNet_v4 - MitoNet_v5 - MitoNet_v6

Reference:

  • doi: 10.1016/j.cels.2022.12.006

dacapo.blockwise.empanada_function.tracker_consensus(trackers, model_config, pixel_vote_thr=2, cluster_iou_thr=0.75, allow_one_view=False, min_size=200, min_extent=4, dtype=np.uint32)

Calculate the orthoplane consensus from trackers. Yields a numpy or zarr volume along with the name of the class that is segmented.

Parameters:

  • trackers (dict) – The trackers dictionary.

  • model_config (str) – The model configuration to use.

  • pixel_vote_thr (int) – The pixel vote threshold.

  • cluster_iou_thr (float) – The cluster IoU threshold.

  • allow_one_view (bool) – Whether to allow one view.

  • min_size (int) – The minimum size of objects.

  • min_extent (int) – The minimum extent of objects.

  • dtype (type) – The data type.

Returns:

The generator object.

Return type:

generator

Raises:

ImportError – If empanada-napari is not installed.

Examples

>>> for vol, class_name, tracker in tracker_consensus(trackers, model_config):
...     print(vol.shape, class_name, tracker)

Note

The model_config parameter should be one of the following: - MitoNet_v1 - MitoNet_v2 - MitoNet_v3 - MitoNet_v4 - MitoNet_v5 - MitoNet_v6

Reference:

  • doi: 10.1016/j.cels.2022.12.006