baby.morph_thresh_seg.MorphSegGrouped

class baby.morph_thresh_seg.MorphSegGrouped(flattener, cellgroups=None, interior_threshold=0.5, nclosing=0, nopening=0, connectivity=2, min_area=10, pedge_thresh=None, fit_radial=False, use_group_thresh=False, group_thresh_expansion=0.0, edge_sub_dilations=None, containment_thresh=0.8, containment_func=<function mask_containment>, return_masks=False, return_coords=False, return_volume=False)

Bases: object

Methods

contains(a, b)

extract_edges(pred, shape, refine_outlines, ...)

remove_duplicates()

Resolve any cells duplicated across adjacent groups:

segment(pred[, refine_outlines, return_volume])

Take the output of the neural network and turn it into an instance segmentation output.
__init__(flattener, cellgroups=None, interior_threshold=0.5, nclosing=0, nopening=0, connectivity=2, min_area=10, pedge_thresh=None, fit_radial=False, use_group_thresh=False, group_thresh_expansion=0.0, edge_sub_dilations=None, containment_thresh=0.8, containment_func=<function mask_containment>, return_masks=False, return_coords=False, return_volume=False)
Parameters

  • flattener –

  • cellgroups –

  • interior_threshold –

  • nclosing –

  • nopening –

  • connectivity –

  • min_area –

  • pedge_thresh –

  • fit_radial –

  • use_group_thresh –

  • group_thresh_expansion –

  • edge_sub_dilations –

  • containment_thresh –

  • containment_func –

  • return_masks –

  • return_coords –

Methods

__init__(flattener[, cellgroups, ...])

param flattener

contains(a, b)

extract_edges(pred, shape, refine_outlines, ...)

remove_duplicates()

Resolve any cells duplicated across adjacent groups:

segment(pred[, refine_outlines, return_volume])

Take the output of the neural network and turn it into an instance segmentation output.
remove_duplicates()

Resolve any cells duplicated across adjacent groups:

Parameters

group_segs –

Returns

The group segmentations with duplicates removed

segment(pred, refine_outlines=False, return_volume=False)

Take the output of the neural network and turn it into an instance segmentation output.

Parameters

pred – list of prediction images (ndarray with shape (x, y))

matching self.flattener.names() :return: a list of boolean edge images (ndarray shape (x, y)), one for each cell identified. If return_masks and/or return_coords are true, the output will be a tuple of edge images, filled masks, and/or radial coordinates.