viper.comparison.distances
Class AttrMeasure

java.lang.Object
  viper.comparison.distances.AttrMeasure

Direct Known Subclasses:

FramewiseEvaluation.FrameMeasure, TrackingEvaluation.TrackingMeasure

public class AttrMeasure

extends java.lang.Object

Associates distance and metric with a specific type of attribute for a specific descriptor to Object Analysis type evaluation.

Constructor Summary

AttrMeasure(Distance metric, double tol)
Constructs a new measure with the given metric and localization threshold, with the default statistic and statistic tolerance.

AttrMeasure(java.lang.String attr)
Gets the default measure for the attribute.

AttrMeasure(java.lang.String attr, java.util.StringTokenizer st)
Constructs a new measure by parsing the measure in text/epf format.

Method Summary

FrameSpan	calculateDistancesAndThresh(Attribute target, FrameSpan targetSpan, Attribute candidate, FrameSpan candSpan, CanonicalFileDescriptor cfd, DistanceHolder dh) Calculate the distances between the two attributes using this attribute measure.
double	distanceAgainst(Measurable.Difference D) Compares attributes or framespans by the metric this is set to use.
double	distanceAgainstDynamic(Attribute target, FrameSpan gSpan, Attribute candidate, FrameSpan fSpan, CanonicalFileDescriptor cfd) compares two dynamic attributes
boolean	equals(java.lang.Object o) Tests to see if this is the same kind of measure as the parameter.
DistanceHolder	getDistances() Gets the current distances.
Distance	getMetric() Gets the metric.
double[]	getStatistics() Gets the current statistics
double	getStatThreshold() Gets the threshold to apply on the given statstic to count as successfuly statisticked.
int	getStatType() Gets the statistic type associated with the measure, e.g.
double	getTolerance() Gets the localization tolerance.
int	hashCode() Computes a hashcode on the metric, statistic, and thresholds.
boolean	isValidFor(java.lang.String attr) Tests that this measure is valid for a given attribute.
static boolean	isValidMeasure(java.lang.String attr, AttrMeasure meas) Tests to see if the given measure is valid for the given attribute type.
static boolean	isValidTolerance(double value) Tests to see if the tolerance is a valid number.
void	setMetric(Distance M) Sets the metric.
void	setTolerance(double tol) Sets the localization tolerance.
boolean	thresh(double value) Determines if the given distance value is less than the current tolerance.
java.lang.String	toString() Prints out the metric and threshold, without the brackets.

Methods inherited from class java.lang.Object

clone, finalize, getClass, notify, notifyAll, wait, wait, wait

Constructor Detail

AttrMeasure

public AttrMeasure(Distance metric,
                   double tol)

Constructs a new measure with the given metric and localization threshold, with the default statistic and statistic tolerance.

Parameters:

metric - the metric

tol - the localization threshold

AttrMeasure

public AttrMeasure(java.lang.String attr)

Gets the default measure for the attribute.

Parameters:

attr - the attribute type

Throws:

java.lang.IllegalStateException - if the default metric is null

AttrMeasure

public AttrMeasure(java.lang.String attr,
                   java.util.StringTokenizer st)
            throws ImproperMetricException

Constructs a new measure by parsing the measure in text/epf format. For example: [ metric threshold ].

Parameters:

attr - the attribute to apply to

st - the text to parse

Throws:

ImproperMetricException - if there is a parsing error or a metric is not found

Method Detail

getStatType

public int getStatType()

Gets the statistic type associated with the measure, e.g. Distances.MEDIAN.

Returns:

the statistic

getStatThreshold

public double getStatThreshold()

Gets the threshold to apply on the given statstic to count as successfuly statisticked.

Returns:

the statistic threshold, for the value of appropriate statistic type for the per-frame distances

equals

public boolean equals(java.lang.Object o)

Tests to see if this is the same kind of measure as the parameter.

Parameters:

o - the measure to test against

Returns:

true if they have the same metric, statistic, and localization and statistic thresholds

hashCode

public int hashCode()

Computes a hashcode on the metric, statistic, and thresholds.

Returns:

a hash code

toString

public java.lang.String toString()

Prints out the metric and threshold, without the brackets. Threshold is either a number or the string (INFINITY)

Returns:

metric threshold

setTolerance

public final void setTolerance(double tol)

Sets the localization tolerance.

Parameters:

tol - the new localization tolerance

getTolerance

public final double getTolerance()

Gets the localization tolerance.

Returns:

the localization tolerance

setMetric

public final void setMetric(Distance M)

Sets the metric.

Parameters:

M - the new metric

getMetric

public final Distance getMetric()

Gets the metric.

Returns:

the metric

thresh

public boolean thresh(double value)

Determines if the given distance value is less than the current tolerance. It deals with the case that the distance may be a score (larger = better) or a distance (smaller = better). Although I generally prefer distances, there will be people who wish to do it the other way.

Parameters:

value - the distance value (or score)

Returns:

if the value is considered good enough to count as a localization

calculateDistancesAndThresh

public FrameSpan calculateDistancesAndThresh(Attribute target,
                                             FrameSpan targetSpan,
                                             Attribute candidate,
                                             FrameSpan candSpan,
                                             CanonicalFileDescriptor cfd,
                                             DistanceHolder dh)
                                      throws MethodNotSupportedException

Calculate the distances between the two attributes using this attribute measure. It returns the set of frames that were successfully localized.

Parameters:

target - the target attribute

targetSpan - the framespan of the target attribute's descriptor

candidate - the candidate attribute

candSpan - the framespan of the candidate attribute's parent descriptor

cfd - information about the media file

dh - this will hold the computed distance information

Returns:

the frames that qualify as localized

Throws:

MethodNotSupportedException - if there is an error in the data types or in the metric

distanceAgainst

public double distanceAgainst(Measurable.Difference D)
                       throws MethodNotSupportedException

Compares attributes or framespans by the metric this is set to use.

Parameters:

D - the difference comparison

Returns:

the value of the metric's Distance.getDistance(viper.descriptors.attributes.Measurable.Difference)method.

Throws:

MethodNotSupportedException

getDistances

public DistanceHolder getDistances()

Gets the current distances.

Returns:

the distances from the last call to calculateDistancesAndThresh(Attribute, FrameSpan, Attribute, FrameSpan, CanonicalFileDescriptor, DistanceHolder).

getStatistics

public double[] getStatistics()

Gets the current statistics

Returns:

the statistics from the last call to calculateDistancesAndThresh(Attribute, FrameSpan, Attribute, FrameSpan, CanonicalFileDescriptor, DistanceHolder).

distanceAgainstDynamic

public double distanceAgainstDynamic(Attribute target,
                                     FrameSpan gSpan,
                                     Attribute candidate,
                                     FrameSpan fSpan,
                                     CanonicalFileDescriptor cfd)
                              throws MethodNotSupportedException

compares two dynamic attributes

Parameters:

target - the target attribute

gSpan - the target attribute span

candidate - the candidate attribute

fSpan - the candidate span

cfd - the source media description

Returns:

the distance

Throws:

MethodNotSupportedException

isValidFor

public boolean isValidFor(java.lang.String attr)

Tests that this measure is valid for a given attribute. Of course, it might not work, anyway, as it only goes by metric name and not implementation.

Parameters:

attr - the attribute type to see if this metric is valid for

Returns:

true if this measure is defined for the specified attribute type

isValidMeasure

public static boolean isValidMeasure(java.lang.String attr,
                                     AttrMeasure meas)

Tests to see if the given measure is valid for the given attribute type.

Parameters:

attr - the attribute type

meas - the measure

Returns:

true when the measure works

isValidTolerance

public static boolean isValidTolerance(double value)

Tests to see if the tolerance is a valid number.

Parameters:

value - the tolerance

Returns:

if the number is between zero and positive infinity