API Reference [update in progress]

platerecipy.model

File brief

model.py

Module for model object and functions.

This is a part of platerecipy package. For license and citation, please refer to the main repository: github.com/pjavaheri/platerecipy

Author(s): Pejvak Javaheri; pejvak.javaheri@mail.utoronto.ca

PlateModel

Bases: object

A class for defining plate detection parameters, field stacking, and performing segmentation.

use_spherical_distance property

Whether to apply use the great circle angle of separation instead of planar distance transform (i.e., for a SphericalGrid).

wraparound_azimuthally property

Whether to apply wraparound boundary conditions along phi (i.e., for a SphericalGrid).

__init__(grid)

Create a PlateModel object by initializing it using a Grid.

Parameters:

Name	Type	Description	Default
grid	(Grid,)	An instance of a grid	required

clear_stacked_field()

To reset and clear the stacked field.

find_plates(boundary_quantile=0.9, boundary_absolute=1.0, separation_tolerance=None, num_threads=1, min_marker_size=None, preserve_small_markers=False, manual_markers=None, identify_nonconforming=False, RW_beta=100.0, RW_solver_tolerance=0.001, RW_solver='LU', return_IDs=True)

Applies segmentation on stacked_field and returns an integer array of the same shape with each cell carrying the plate ID (i.e., segment number). This function can be called multiple times from the same object instance as it does not alter stacked_field.

Warning

There are several options that fundamentally alter the way this function works. The user is suggested to carefully read the documentation for all input parameters regardless of their apparent relevance to the specific task at hand.

Warning

If both boundary_quantile and boundary_absolute are specified, a node qualifies as a possible boundary point so long as it satisfies one of the two conditions.

Parameters:

Name	Type	Description	Default
boundary_quantile	float	The quantile that represents values above which that are presumed to feature plate boundaries.	0.9
boundary_absolute	float	The absolute value that represents values above which that are presumed to feature plate boundaries.	1.0
separation_tolerance	float	Whether to use a distance transform and consider regions with distances less than separation_tolerance from the boundary (determined by boundary_quantile) as a part of the boundary as well. This option is useful when segments (plates) have imperfect boundaries that require some separation tolerance to close them. For a SphericalGrid, it is presumed that the field is a mercator projection and instead of a Euclidean distance transform, a spherical one will be applied. In that case, separation_tolerance will be considered as the angle of tolerance on the great circle passing through a given pair of points in radians.	None
num_threads	int	Number of threads to use to perform the spherical distance transform.	1
min_marker_size	int	If provided, watershed markers will be filtered such that markers with fewer cells/pixels than min_marker_size will be ignored. This is useful when then input field is noisy or not coherent enough.	None
preserve_small_markers	bool	Whether to reinstate small markers obscured by by the separation tolerance step.	False
manual_markers	ndarray	This option ignores all previous arguments relating to marker specification. Instead, it allows the user to manually provide a 2D Boolean array of the same shape as self.stacked_field that indicates the seeds to the RW algorithm.	None
identify_non_conforming	bool	Whether to extract and separately label non-conforming regions defined by featuring a stacked field greater than 0.5 and a ID probability less than 0.5 with ID=0.	False
RW_beta	float	Gaussian beta parameter for random walker connection weights. This parameter controls the sharpness of the boundaries.	100.
RW_solver_tolerance	float	Tolerance value for the choice of solver that requires a tolerance (e.g., if RW_solver = 'CG').	1e-3
RW_solver	str	Numerical solver used to obtain an random walker probability solution. Possible choices: 'direct', 'LU', 'CG', or 'FA'.	'LU'
return_IDs	bool	Whether to return a copy of self.plate_IDs.	True

stack_field(field, invert=False, take_log=False, stack_weight=1.0)

Stack a new field by normalizing the field values.

Parameters:

Name	Type	Description	Default
field	(ndarray,)	The field to be stacked.	required
invert	bool	If increase in input field corresponds to a decrease in deformation.	False,
take_log	bool	If the input field varies by orders of magnitude.	False,
stack_weight	float	The corresponding weight when stacked (between 0 and 1).	1.,