net.imglib2.algorithm.metrics.segmentation

Class ConfusionMatrix<I extends net.imglib2.type.numeric.IntegerType<I>,J extends net.imglib2.type.numeric.IntegerType<J>>

java.lang.Object

net.imglib2.algorithm.metrics.segmentation.ConfusionMatrix<I,J>

Type Parameters:

I - Ground-truth pixel type

J - Predicted pixel type

public class ConfusionMatrix<I extends net.imglib2.type.numeric.IntegerType<I>,J extends net.imglib2.type.numeric.IntegerType<J>>
extends Object

In an instance segmentation problem, where each instance is represented by an ensemble of pixels with equal value, the confusion matrix between the ground-truth and a predicted image is a 2D matrix whose elements Cij are equal to the number of pixels shared between ground-truth label i and predicted label j.

Additionally, a ConfusionMatrix gives access to the total number of pixels labeled by ground-truth label i or predicted label j.

Pixel of value 0 are considered background and ignored in confusion matrix. Therefore, if ground-truth image has M labels, including background, and the prediction image has N, then the confusion matrix is of size (M-1)x(N-1).

Constructor Summary

Constructors

Constructor and Description
ConfusionMatrix(net.imglib2.RandomAccessibleInterval<I> groundTruth, net.imglib2.RandomAccessibleInterval<J> prediction) Constructor.

Method Summary

Modifier and Type	Method and Description
int	getGroundTruthLabelSize(int labelIndex) Return the number of pixels corresponding to the ground-truth label indexed by labelIndex.
int	getIntersection(int gtLabelIndex, int predLabelIndex) Return the number of pixels shared by the ground-truth label indexed by gtIndex and the prediction label indexed by predIndex.
int	getNumberGroundTruthLabels() Return the number of ground-truth labels.
int	getNumberPredictionLabels() Return the number of prediction labels.
int	getPredictionLabelSize(int labelIndex) Return the number of pixels corresponding to the prediction label indexed by labelIndex.

Methods inherited from class java.lang.Object

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

Constructor Detail

ConfusionMatrix

public ConfusionMatrix(net.imglib2.RandomAccessibleInterval<I> groundTruth,
                       net.imglib2.RandomAccessibleInterval<J> prediction)

Constructor. Compute the histogram for all ground-truth and predicted labels, then the 2D array representing the confusion matrix.

Parameters:

groundTruth - Ground-truth image

prediction - Prediction image

Method Detail

getGroundTruthLabelSize

public int getGroundTruthLabelSize(int labelIndex)

Return the number of pixels corresponding to the ground-truth label indexed by labelIndex.

Parameters:

labelIndex - Index of the label

Returns:

Number of pixels, or -1 if the index is out of bounds

getPredictionLabelSize

public int getPredictionLabelSize(int labelIndex)

Return the number of pixels corresponding to the prediction label indexed by labelIndex.

Parameters:

labelIndex - Index of the label

Returns:

Number of pixels, or -1 if the index is out of bounds

getIntersection

public int getIntersection(int gtLabelIndex,
                           int predLabelIndex)

Return the number of pixels shared by the ground-truth label indexed by gtIndex and the prediction label indexed by predIndex.

Parameters:

gtLabelIndex - Index of the ground-truth label

predLabelIndex - Index of the prediction label

Returns:

Number of pixels shared by the two labels

getNumberGroundTruthLabels

public int getNumberGroundTruthLabels()

Return the number of ground-truth labels.

Returns:

Number of ground-truth labels

getNumberPredictionLabels

public int getNumberPredictionLabels()

Return the number of prediction labels.

Returns:

Number of prediction labels