org.openimaj.ml.annotation.basic

Class KNNAnnotator<OBJECT,ANNOTATION,FEATURE>

java.lang.Object

org.openimaj.ml.annotation.AbstractAnnotator<OBJECT,ANNOTATION>

org.openimaj.ml.annotation.IncrementalAnnotator<OBJECT,ANNOTATION>

org.openimaj.ml.annotation.basic.KNNAnnotator<OBJECT,ANNOTATION,FEATURE>

Type Parameters:

OBJECT - Type of object being annotated

ANNOTATION - Type of annotation

FEATURE - Type of feature produced by extractor

All Implemented Interfaces:

Classifier<ANNOTATION,OBJECT>, Annotator<OBJECT,ANNOTATION>, IncrementalTrainer<Annotated<OBJECT,ANNOTATION>>

public class KNNAnnotator<OBJECT,ANNOTATION,FEATURE>
extends IncrementalAnnotator<OBJECT,ANNOTATION>

Annotator based on a multi-class k-nearest-neighbour classifier. Uses a ObjectNearestNeighboursExact to perform the kNN search, so is applicable to any objects that can be compared with a DistanceComparator.

Author:

Jonathon Hare (jsh2@ecs.soton.ac.uk)

Field Summary

Fields

Modifier and Type	Field and Description
protected List<Collection<ANNOTATION>>	annotations
protected Set<ANNOTATION>	annotationsSet
protected DistanceComparator<? super FEATURE>	comparator
protected FeatureExtractor<FEATURE,OBJECT>	extractor
protected List<FEATURE>	features
protected int	k
protected ObjectNearestNeighbours<FEATURE>	nn
protected float	threshold

Constructor Summary

Constructors

Constructor and Description
KNNAnnotator(FeatureExtractor<FEATURE,OBJECT> extractor, DistanceComparator<? super FEATURE> comparator) Construct with the given extractor and comparator.
KNNAnnotator(FeatureExtractor<FEATURE,OBJECT> extractor, DistanceComparator<? super FEATURE> comparator, float threshold) Construct with the given extractor, comparator and threshold.
KNNAnnotator(FeatureExtractor<FEATURE,OBJECT> extractor, DistanceComparator<? super FEATURE> comparator, int k) Construct with the given extractor, comparator and number of neighbours.
KNNAnnotator(FeatureExtractor<FEATURE,OBJECT> extractor, DistanceComparator<? super FEATURE> comparator, int k, float threshold) Construct with the given extractor, comparator, number of neighbours and threshold.

Method Summary

Modifier and Type	Method and Description
List<ScoredAnnotation<ANNOTATION>>	annotate(OBJECT object) Generate annotations for the given object.
static <OBJECT,ANNOTATION,EXTRACTOR extends FeatureExtractor<FEATURE,OBJECT>,FEATURE> KNNAnnotator<OBJECT,ANNOTATION,FEATURE>	create(EXTRACTOR extractor, DistanceComparator<FEATURE> comparator) Create a new KNNAnnotator with the given extractor and comparator.
static <OBJECT,ANNOTATION,EXTRACTOR extends FeatureExtractor<FEATURE,OBJECT>,FEATURE> KNNAnnotator<OBJECT,ANNOTATION,FEATURE>	create(EXTRACTOR extractor, DistanceComparator<FEATURE> comparator, float threshold) Create a new KNNAnnotator with the given extractor, comparator and threshold.
static <OBJECT,ANNOTATION,EXTRACTOR extends FeatureExtractor<FEATURE,OBJECT>,FEATURE> KNNAnnotator<OBJECT,ANNOTATION,FEATURE>	create(EXTRACTOR extractor, DistanceComparator<FEATURE> comparator, int k) Create a new KNNAnnotator with the given extractor, comparator and number of neighbours.
static <OBJECT,ANNOTATION,EXTRACTOR extends FeatureExtractor<FEATURE,OBJECT>,FEATURE> KNNAnnotator<OBJECT,ANNOTATION,FEATURE>	create(EXTRACTOR extractor, DistanceComparator<FEATURE> comparator, int k, float threshold) Create a new KNNAnnotator with the given extractor, comparator, number of neighbours and threshold.
Set<ANNOTATION>	getAnnotations()
int	getK()
void	reset() Reset the object to its initial condition, as if it hasn't seen any training data.
void	setK(int k) Set the number of neighbours
void	train(Annotated<OBJECT,ANNOTATION> annotated) Train/update object using a new instance.

Methods inherited from class org.openimaj.ml.annotation.IncrementalAnnotator

train, train, trainMultiClass

Methods inherited from class org.openimaj.ml.annotation.AbstractAnnotator

classify

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Detail

k

protected int k

features

protected final List<FEATURE> features

annotations

protected final List<Collection<ANNOTATION>> annotations

annotationsSet

protected final Set<ANNOTATION> annotationsSet

nn

protected ObjectNearestNeighbours<FEATURE> nn

comparator

protected DistanceComparator<? super FEATURE> comparator

threshold

protected final float threshold

extractor

protected FeatureExtractor<FEATURE,OBJECT> extractor

Constructor Detail

KNNAnnotator

public KNNAnnotator(FeatureExtractor<FEATURE,OBJECT> extractor,
                    DistanceComparator<? super FEATURE> comparator,
                    float threshold)

Construct with the given extractor, comparator and threshold. The number of neighbours is set to 1.

If the comparator defines a distance, then only scores below the distance will be accepted. If the threshold defines a similarity, then only scores above the threshold will be accepted.

Parameters:

extractor - the extractor

comparator - the comparator

threshold - the threshold for successful matches

KNNAnnotator

public KNNAnnotator(FeatureExtractor<FEATURE,OBJECT> extractor,
                    DistanceComparator<? super FEATURE> comparator)

Construct with the given extractor and comparator. The number of neighbours is set to 1. The threshold test is disabled.

Parameters:

extractor - the extractor

comparator - the comparator

KNNAnnotator

public KNNAnnotator(FeatureExtractor<FEATURE,OBJECT> extractor,
                    DistanceComparator<? super FEATURE> comparator,
                    int k)

Construct with the given extractor, comparator and number of neighbours. The distance threshold is disabled.

Parameters:

extractor - the extractor

comparator - the comparator

k - the number of neighbours

KNNAnnotator

public KNNAnnotator(FeatureExtractor<FEATURE,OBJECT> extractor,
                    DistanceComparator<? super FEATURE> comparator,
                    int k,
                    float threshold)

Construct with the given extractor, comparator, number of neighbours and threshold.

If the comparator defines a distance, then only scores below the distance will be accepted. If the threshold defines a similarity, then only scores above the threshold will be accepted.

Parameters:

extractor - the extractor

comparator - the comparator

k - the number of neighbours

threshold - the threshold on distance for successful matches

Method Detail

create

public static <OBJECT,ANNOTATION,EXTRACTOR extends FeatureExtractor<FEATURE,OBJECT>,FEATURE> KNNAnnotator<OBJECT,ANNOTATION,FEATURE> create(EXTRACTOR extractor,
                                                                                                                                            DistanceComparator<FEATURE> comparator,
                                                                                                                                            float threshold)

Create a new KNNAnnotator with the given extractor, comparator and threshold. The number of neighbours is set to 1.

If the comparator defines a distance, then only scores below the distance will be accepted. If the threshold defines a similarity, then only scores above the threshold will be accepted.

Type Parameters:

OBJECT - Type of object being annotated

ANNOTATION - Type of annotation

EXTRACTOR - Type of feature extractor

FEATURE - Type of feature produced by extractor

Parameters:

extractor - the extractor

comparator - the comparator

threshold - the threshold for successful matches

Returns:

new KNNAnnotator

create

public static <OBJECT,ANNOTATION,EXTRACTOR extends FeatureExtractor<FEATURE,OBJECT>,FEATURE> KNNAnnotator<OBJECT,ANNOTATION,FEATURE> create(EXTRACTOR extractor,
                                                                                                                                            DistanceComparator<FEATURE> comparator)

Create a new KNNAnnotator with the given extractor and comparator. The number of neighbours is set to 1. The threshold test is disabled.

Type Parameters:

OBJECT - Type of object being annotated

ANNOTATION - Type of annotation

EXTRACTOR - Type of feature extractor

FEATURE - Type of feature produced by extractor

Parameters:

extractor - the extractor

comparator - the comparator

Returns:

new KNNAnnotator

create

public static <OBJECT,ANNOTATION,EXTRACTOR extends FeatureExtractor<FEATURE,OBJECT>,FEATURE> KNNAnnotator<OBJECT,ANNOTATION,FEATURE> create(EXTRACTOR extractor,
                                                                                                                                            DistanceComparator<FEATURE> comparator,
                                                                                                                                            int k)

Create a new KNNAnnotator with the given extractor, comparator and number of neighbours. The distance threshold is disabled.

Type Parameters:

OBJECT - Type of object being annotated

ANNOTATION - Type of annotation

EXTRACTOR - Type of feature extractor

FEATURE - Type of feature produced by extractor

Parameters:

extractor - the extractor

comparator - the comparator

k - the number of neighbours

Returns:

new KNNAnnotator

create

public static <OBJECT,ANNOTATION,EXTRACTOR extends FeatureExtractor<FEATURE,OBJECT>,FEATURE> KNNAnnotator<OBJECT,ANNOTATION,FEATURE> create(EXTRACTOR extractor,
                                                                                                                                            DistanceComparator<FEATURE> comparator,
                                                                                                                                            int k,
                                                                                                                                            float threshold)

Create a new KNNAnnotator with the given extractor, comparator, number of neighbours and threshold.

If the comparator defines a distance, then only scores below the distance will be accepted. If the threshold defines a similarity, then only scores above the threshold will be accepted.

Type Parameters:

OBJECT - Type of object being annotated

ANNOTATION - Type of annotation

EXTRACTOR - Type of feature extractor

FEATURE - Type of feature produced by extractor

Parameters:

extractor - the extractor

comparator - the comparator

k - the number of neighbours

threshold - the threshold on distance for successful matches

Returns:

new KNNAnnotator

train

public void train(Annotated<OBJECT,ANNOTATION> annotated)

Description copied from interface: IncrementalTrainer

Train/update object using a new instance.

Parameters:

annotated - instance to train with

reset

public void reset()

Description copied from interface: IncrementalTrainer

Reset the object to its initial condition, as if it hasn't seen any training data.

getAnnotations

public Set<ANNOTATION> getAnnotations()

Returns:

a Set of all annotations this Annotator knows about

annotate

public List<ScoredAnnotation<ANNOTATION>> annotate(OBJECT object)

Description copied from interface: Annotator

Generate annotations for the given object.

Parameters:

object - the image

Returns:

generated annotations

getK

public int getK()

Returns:

the number of neighbours to search for

setK

public void setK(int k)

Set the number of neighbours

Parameters:

k - the number of neighbours