Classification

Last modified: February 14, 2023

Contents

classify_automatic

classify_automatic (str id)

Operates on:Image [OneBit]
Category:Classification
Defined in:plugin.py
Author:Michael Droettboom and Karl MacMillan

Classifies the image as the value id_name and sets the state to AUTOMATIC. Use this method when an automatic classifier has classified this glyph. Note that this method does not touch the confidence map for the main id; this must be set separately.

id_name

Can come in one of two forms:

string
image is classified using the given .-delimited class name. As in this case no confidence is provided, it is set to zero.
list of tuples

A list of tuples where each tuple is the pair (confidence, class_name).

confidence
A value in range (0, 1), where 0 is uncertain and 1 is certain.
class_name
A .-delimited class name.

classify_heuristic

classify_heuristic (str id)

Operates on:Image [OneBit]
Category:Classification
Defined in:plugin.py
Author:Michael Droettboom and Karl MacMillan

Classifies the image as the value id_name and sets the state to HEURISTIC. Use this method when a heuristic process has classified this glyph. This method also unsets the confidence map.

id_name

Can come in one of two forms:

string
image is classified using the given .-delimited class name. As in this case no confidence is provided, it is set to 0.5.
list of tuples

A list of tuples where each tuple is the pair (confidence, class_name).

confidence
A value in range (0, 1), where 0 is uncertain and 1 is certain.
class_name
A .-delimited class name.

classify_manual

classify_manual (str id)

Operates on:Image [OneBit]
Category:Classification
Defined in:plugin.py
Author:Michael Droettboom and Karl MacMillan

Classifies the image as the value id_name and sets the state to MANUAL. Use this method when an end user has classified this glyph. This method also unsets the confidence map.

id_name

Can come in one of two forms:

string
image is classified using the given .-delimited class name. As in this case no confidence is provided, it is set to one.
list of tuples

A list of tuples where each tuple is the pair (confidence, class_name).

confidence
A value in range (0, 1), where 0 is uncertain and 1 is certain.
class_name
A .-delimited class name.

get_confidence

float get_confidence (Choice [] confidence_type = None)

Operates on:Image [OneBit]
Returns:float
Category:Classification
Defined in:plugin.py
Author:Michael Droettboom and Karl MacMillan

Returns the confidence of the primary class assigned to the image. confidence_type can be one of the predefined confidence measures, e.g.

from gamera.gameracore import CONFIDENCE_KNNFRACTION
glyph.get_confidence(CONFIDENCE_KNNFRACTION)

When no confidence_type is given, the default confidence stored with id_name[0] is returned. If the image is unclassified, -1.0 is returned.

For details about the confidence computation in Gamera, see

C. Dalitz: Reject Options and Confidence Measures for kNN Classifiers. In C. Dalitz (Ed.): "Document Image Analysis with the Gamera Framework." Schriftenreihe des Fachbereichs Elektrotechnik und Informatik, Hochschule Niederrhein, vol. 8, pp. 16-38, Shaker Verlag (2009).

get_main_id

str get_main_id ()

Operates on:Image [OneBit]
Returns:str
Category:Classification
Defined in:plugin.py
Author:Michael Droettboom and Karl MacMillan

Returns the primary class name assigned to the image.

If the image is unclassified, the result is 'UNCLASSIFIED'.

has_id_name

bool has_id_name (str id)

Operates on:Image [OneBit]
Returns:bool
Category:Classification
Defined in:plugin.py
Author:Michael Droettboom and Karl MacMillan

Returns True if the image has the given classification name.

match_id_name

match_id_name (str regex)

Operates on:Image [OneBit|GreyScale|Grey16|RGB|Float|Complex]
Category:Classification
Defined in:id_name_matching.py
Author:Michael Droettboom and Karl MacMillan

Returns true if the image's main id_name matches the given regular expression.

regex
A special-purpose regular expression as defined below.

A special-purpose regular expression language designed for convenience on Gamera symbol names is used here. It has limited flexibility relative to Python regular expressions, and is more akin to wildcards in command-line shells. This language is converted to standard Python regular expressions (re module) for efficiency.

Informal syntax definition:

Regular expression Description
A|B matches A or B
A.B|C matches A.B or A.C
* multiple-character wildcard
? single character wildcard
() grouping can be performed with parentheses
[a-z] matches any character a-z

Example expressions:

(upper.x)|(lower.y) match either upper.x or lower.y
upper.x|upper.y careful! matches upper.x.y or upper.upper.y
upper.* match anything in the upper category
upper.a|b|c matches upper.a, upper.b or upper.c
upper.capital_? ? is a single character wildcard

unclassify

unclassify ()

Operates on:Image [OneBit]
Category:Classification
Defined in:plugin.py
Author:Michael Droettboom and Karl MacMillan

Sets the image back to an UNCLASSIFIED state. Use this when you are not longer sure of the identity of the image and you want an automatic classifier to reclassify.