Contents

API

Labeling

class lwsspy.ml.labeling.segmentlabeler.SegmentLabeler(img, segmentation, labeldict={'410': 2, '660': 3, 'moho': 1, 'none': 9999}, loglevel=20)[source]
__init__(img, segmentation, labeldict={'410': 2, '660': 3, 'moho': 1, 'none': 9999}, loglevel=20)[source]

This class takes in an image, some corresponding segmentation e.g. (SLIC) and a label dictionary that can be used to label the image in an ‘intuitive’ GUI.

Parameters
imgndarray [w x h x 3]

Image

segmentationndarray [w x h]

mask that has a unique nunmber for each segment such that it can be labeled.

labeldictdict, optional

Dictionary of labels, must contain the ‘none’ keyword, which denotes the unlabeled value, by default {‘moho’: 1, ‘410’: 2, ‘660’: 3, ‘none’: 9999}

loglevellogging.LOGLEVEL, optional

loglevel, used to debug the event loop. Not necessary to be modified, by default logging.INFO

Notes

Authors

Lucas Sawade (lsawade@princeton.edu)

Last Modified

2021.07.02 00.00 (Lucas Sawade)

Examples

The usage of this labeling is fairly straight forward. Given an image, img, and a segmentation, segs, of said image, we instantiate the class and call the start labeling method.

>>> from lwsspy.ml import SegmentLabeler
>>> sl = SegmentLabeler(img, segments)
>>> labeled_mask = sl.start_labeling()

This will open two figures. One contains the image by itself and the second one contains the image and the outlines of the mask. We will use the firs image for reference and the second image to actually label the image segments. The GUI has a few controls that can be used.

Control

Action

Mouse-left

Add label to segment

Mouse-right

Remove label to segment

Mouse-left-drag

Add labels to segments dragged over

Mouse-right-drag

Remove labels to segments dragged over

n

Next label

p

Previous label True

d

Delete previously labeled segment

esc

Close figure and return the currently selected mask

The currently selected mask will also be returned if any of the figures is closed.

The selected mask can then be viewed via

>>> import matplotlib.pyplot as plt
>>> imshow(labeled_mask, aspect='auto')

Note that depending on the values you chose in the label dictionary you will have to create a colormap and norm that resembles the mask values.

class lwsspy.ml.labeling.volumelabeler.VolumeLabeler(x, y, z, V, labeldict={'410': 2, '660': 3, 'moho': 1, 'none': 9999}, labeled: Optional[dict] = None)[source]
__init__(x, y, z, V, labeldict={'410': 2, '660': 3, 'moho': 1, 'none': 9999}, labeled: Optional[dict] = None)[source]

This class uses the :class:lwsspy.ml.labeling.segementlabeler.SegmentLabeler to label slices in a volume. After instantiation, use the label method to label a Volumen of single Channel data.

Parameters
xarraylike

x vector

yarralike

y vector

zarraylike

z vector

Varraylike

3D array with data corresponding to vectors X x Y x Z direction.

labeldictdict, optional

dictionary with labels, by default {‘moho’: 1, ‘410’: 2, ‘660’: 3, ‘none’: 9999}

labeledOptional[dict], optional

Optional dictionary with labeled volume. Labeled should contain 4 variables lx, ly, lz, lV, where l{x,y,z} are boolean arrays that where defined as locations for labeled layers. Useful if you need to take a labeling break. Right now there is no way of resuming to an image make sure you finisshe the ones you have started, by default None

Notes

Things that I have to fix:

  1. The volume should be allowed to be 3-channel data

  2. Also allow for custom normalizations and colormaps

Authors

Lucas Sawade (lsawade@princeton.edu)

Last Modified

2021.07.02 00.00 (Lucas Sawade)

Installation