Skip to content

GitLab

M
MultiCube
Commit 7482508b authored by balanche's avatar balanche
Browse files
Options

work in progess : cleaning of the old distance implementation

parent 965162f4
Hide whitespace changes
Inline Side-by-side
Showing with 15 additions and 829 deletions
src/mustic/gui/MainFrame.java
View file @ 7482508b
......@@ -10,12 +10,6 @@ import java.awt.event.*;
import java.awt.image.BufferedImage;
import java.beans.PropertyVetoException;
import java.io.*;
import java.nio.file.DirectoryStream;
import java.nio.file.Files;
import java.nio.file.Path;
import java.nio.file.Paths;
import java.util.ArrayList;
import java.util.Iterator;
import java.util.List;
import java.util.Vector;
import java.util.zip.ZipEntry;
......@@ -27,28 +21,16 @@ import javax.swing.border.EmptyBorder;
import javax.swing.event.*;
import javax.swing.tree.*;
import org.joda.time.DateTime;
import org.joda.time.format.DateTimeFormat;
import org.joda.time.format.DateTimeFormatter;
import jcl.Classification;
import jcl.clustering.Cluster;
import jcl.clustering.ClusteringResult;
import jcl.clustering.constraints.CannotLinkConstraint;
import jcl.clustering.constraints.Constraint;
import jcl.clustering.constraints.MustLinkConstraint;
import jcl.data.Data;
import jcl.data.DataObject;
import jcl.data.SimpleData;
import jcl.data.attribute.AttributeMultiDimSequence;
import jcl.io.results.CSVResultWriter;
import jcl.learning.LearningMethod;
import jcl.learning.methods.monostrategy.SingleClassification;
import jcl.learning.methods.monostrategy.kmeans.ParametersKmeans;
import jcl.learning.methods.multistrategy.samarah.HybridClassification;
import jcl.learning.methods.multistrategy.samarah.SamarahConfig;
import jcl.weights.ClassificationWeights;
import jcl.weights.GlobalWeights;
import jsl.Segmentation;
import mustic.gui.dialog.*;
import mustic.gui.dialog.arff.ArffLoadDialog;
......@@ -68,13 +50,9 @@ import mustic.gui.progressbar.OpenRawImage;
import mustic.gui.segmentation.*;
import mustic.io.*;
import mustic.models.gui.panels.*;
import mustic.models.thread.ClassificationThread;
import mustic.utils.UnZipSub;
import mustic.utils.filters.ImageFileFilter;
import mustic.utils.filters.ZipFileFilter;
import mustic.utils.image.ImageHelper;
import mustic.utils.io.CSVUtils;
import mustic.utils.io.dataExchange.DataCsvExchange;
import mustic.utils.tools.ProgressStatus;
import net.infonode.tabbedpanel.*;
import net.infonode.tabbedpanel.titledtab.TitledTab;
......@@ -736,16 +714,6 @@ public class MainFrame extends JFrame {
});
OGRMenu.add(menuItemLoadOGRToData);
//TODO to delete when no more used
JMenuItem customToDelete = new JMenuItem("custom"); //$NON-NLS-1$
customToDelete.addActionListener(new ActionListener() {
@Override
public void actionPerformed(ActionEvent e) {
customClassify();
}
});
CSVMenu.add(customToDelete);
// menuItemResultToCSV = new JMenuItem(Messages.getString("MainFrame.305")); //$NON-NLS-1$
// menuItemResultToCSV.addActionListener(new ActionListener() {
// @Override
......@@ -3006,14 +2974,6 @@ public class MainFrame extends JFrame {
classifAfterSegmentation = true;
}
else if (entry.getName().equals(Messages.getString("MainFrame.186") + i + "/ParameterMode.ser")) { //$NON-NLS-1$ //$NON-NLS-2$
/**
* Chargement du mode de comparaison des séquences
* @author Francois Petitjean
*/
ObjectInputStream ois = new ObjectInputStream(in);
AttributeMultiDimSequence.setMode((Integer) ois.readObject());
}
else if (entry.getName().equals(Messages.getString("MainFrame.188") + i + "/ParameterSimplify.ser")) { //$NON-NLS-1$ //$NON-NLS-2$
/**
* Chargement du type de simplification des séquences
......@@ -3798,296 +3758,6 @@ public class MainFrame extends JFrame {
this.launchOrthoRect = launchOrthoRect;
}
private void customClassify() {
HybridClassification classification = new HybridClassification();
String datasetName = "FacesUCR";
String datasetPath = "FacesUCR";
String dataPath = "/home/baptiste/A2CNES/";
String resultPath = "/home/baptiste/A2CNES/results/";
String testResultPath = "/home/baptiste/A2CNES/Train_results/";
Data dataTrain = getDataFromFile(dataPath+datasetPath+"/train/"+datasetName+".data", '\t', "train", null);
Data dataTest = getDataFromFile(dataPath+datasetPath+"/test/"+datasetName+".data", '\t', "test", null);
// AttributeMultiDimSequence.setMode(AttributeMultiDimSequence.EUCLIDIEN);
AttributeMultiDimSequence.setMode(AttributeMultiDimSequence.DTW_BARYCENTRE);
DataDesktopFrame[] desktopFrames = MainFrame.getInstance().getDesktop().getAllDataDesktopFrames();
DataSession testSession = desktopFrames[desktopFrames.length-2].getDataSession();
final DateTime startTime = DateTime.now();
int nInf = 12;
int nSup = 17;
// parametre pour la precision des conflits
double minC = 0.9;
// parametre pour les classes representatives
double pcr = 0.2;
// parametre qualite/similitude
double ps = 0.2;
double pq = 1.0 - ps;
double constraintsWgt = 80;
double valueKExtern = 60;
double kExtern = valueKExtern * ((100-constraintsWgt)/100);
double kIntern = (100 - valueKExtern) * ((100-constraintsWgt)/100);
int solutionType = 3;
double classRatio = 0.01;
int criterion = 0;
double degradation = 0.9;
int unificationType = SamarahConfig.WITHOUT_UNCLASSIFIED_UNIFICATION;
classification.setParameters(nInf, nSup, minC, ps, pq, pcr);
classification.setAdvancedParameters(degradation, classRatio, solutionType, kIntern,
kExtern, unificationType, criterion, constraintsWgt);
ClassificationWeights weights = new GlobalWeights(dataTrain);
final Vector<Thread> threadList = new Vector<Thread>();
final Vector<Classification> classifList = new Vector<Classification>();
// we search for all constraints files
// <<<< START FOR PARAM config
// for (int i = 0 ; i < 10 ; i++) {
// HybridClassification classif = (HybridClassification) classification.clone();
// if (i % 2 == 1 ) {
// classification.setParameters(nInf, nSup, minC, 0.4, 0.6, pcr);
// }
// if (i <= 5)
// classif.addAgent(new ParametersKmeans(8, 25, weights), dataTrain);
// if (i <= 7)
// classif.addAgent(new ParametersKmeans(10, 25, weights), dataTrain);
// classif.addAgent(new ParametersKmeans(13, 25, weights), dataTrain);
// classif.addAgent(new ParametersKmeans(16, 25, weights), dataTrain);
// if (i >= 2)
// classif.addAgent(new ParametersKmeans(19, 25, weights), dataTrain);
// if (i >= 4)
// classif.addAgent(new ParametersKmeans(22, 25, weights), dataTrain);
//// classif.addAgent(new ParametersKmeans(3, 10, weights), dataTest);
//// classif.addAgent(new ParametersKmeans(4, 10, weights), dataTest);
//// classif.addAgent(new ParametersKmeans(6, 10, weights), dataTest);
//
// classif.setName(testResultPath+datasetName+"/clusteing"+i);
// classif.setData(dataTrain);
// >>>> END FOR PARAM config1
// for (int i = 0 ; i < 20 ; i++) {
// HybridClassification classif = (HybridClassification) classification.clone();
// classif.addAgent(new ParametersKmeans(13, 20, weights), dataTest);
// classif.addAgent(new ParametersKmeans(16, 20, weights), dataTest);
// classif.addAgent(new ParametersKmeans(19, 20, weights), dataTest);
//
// classif.setName(resultPath+datasetPath+"/clustering"+i);
// classif.setData(dataTest);
//// >>>> REPLACE PARAM
// Thread t = null;
//
// ClassificationImage classificationImage = new ClassificationImage(testSession,
// Messages.getString("ClassifierPanel.73") + DataSession.nbClustering, //$NON-NLS-1$
// true);
//
//
// t = new ClassificationThread(classif, classificationImage.getProgressBar(),
// classificationImage, null);
// t.start();
// try {
// testSession.associatedFrame.setMaximum(true);
// } catch (PropertyVetoException e1) {}
// testSession.addClassifier(classificationImage);
// try {
// testSession.associatedFrame.setSelected(true);
// testSession.associatedFrame.toFront();
// } catch (PropertyVetoException e) {
// e.printStackTrace();
// }
// classificationImage.setVisible(true);
//
// threadList.add(t);
// classifList.add(classif);
// }
// <<<< CONSTRAINTS
try (DirectoryStream<Path> dirStream = Files.newDirectoryStream(
Paths.get(dataPath+datasetName+"/train/"), "*1_10.constraints")) {
Iterator<Path> iter = dirStream.iterator();
while(iter.hasNext()) {
Vector<Constraint> constraints = new Vector<Constraint>();
String filename = iter.next().getFileName().toString();
System.out.println(filename);
CSVUtils reader = null;
try {
reader = new CSVUtils(new File(dataPath+datasetName+"/test/"+filename), '\t');
List<String> line;
while((line = reader.readNext()) != null) {
switch (Integer.parseInt(line.get(line.size()-1))) {
case Constraint.MUST_LINK_TYPE :
try {
int index1 = Integer.parseInt(line.get(0))-1;
int index2 = Integer.parseInt(line.get(1))-1;
constraints.add(new MustLinkConstraint(index1, index2));
} catch(Exception e) {
e.printStackTrace();
}
break;
case Constraint.CANNOT_LINK_TYPE :
try {
int index1 = Integer.parseInt(line.get(0))-1;
int index2 = Integer.parseInt(line.get(1))-1;
constraints.add(new CannotLinkConstraint(index1, index2));
} catch(Exception e) {
e.printStackTrace();
}
break;
}
}
} catch (FileNotFoundException e) {
e.printStackTrace();
}
for(int i = 0 ; i < 1; i++) {
HybridClassification classif = (HybridClassification) classification.clone();
Data currentData = (Data) dataTest.clone();
currentData.updateAndSetConstraintsToSample(constraints);
classif.addAgent(new ParametersKmeans(16, 15, weights), currentData);
classif.addAgent(new ParametersKmeans(20, 15, weights), currentData);
classif.addAgent(new ParametersKmeans(24, 15, weights), currentData);
classif.setName(resultPath+datasetName+"/"+
ImageHelper.stripExtension(filename)+"-"+i+
".clustering");
classif.setData(currentData);
Thread t = null;
ClassificationImage classificationImage = new ClassificationImage(testSession,
classif.getName(), //$NON-NLS-1$
true);
t = new ClassificationThread(classif, classificationImage.getProgressBar(),classificationImage.getStatusBar(),
classificationImage, null, false, null);
t.start();
try {
testSession.associatedFrame.setMaximum(true);
} catch (PropertyVetoException e1) {}
testSession.addClassifier(classificationImage);
try {
testSession.associatedFrame.setSelected(true);
testSession.associatedFrame.toFront();
} catch (PropertyVetoException e) {
e.printStackTrace();
}
classificationImage.setVisible(true);
threadList.add(t);
classifList.add(classif);
}
}
} catch (IOException e2) {
e2.printStackTrace();
}
// >>>> CONSTRAINTS
new Thread() {
@Override
public void run() {
try {
for (Thread thread : threadList) {
thread.join();
}
} catch (InterruptedException e) {
e.printStackTrace();
}
for(Classification cl : classifList) {
try {
new CSVResultWriter(cl, cl.getName()).write();
} catch (IOException e) {
e.printStackTrace();
}
}
DateTimeFormatter formatter = DateTimeFormat.forPattern("MM/dd/yyyy HH:mm:ss");
System.out.println("Start at "+ formatter.print(startTime));
System.out.println("wrote final results at "+ formatter.print(DateTime.now()));
}
}.start();
ClassificationFrame.getInstance().updateClassifierPanel();
ClassificationFrame.getInstance().dispose();
}
private Data getDataFromFile(String path, char separator, String name, String mask) {
Data data = null;
// we create the two data session
// the first session is used for training, the second for classification
try {
List<DataObject> result = new ArrayList<DataObject>();
CSVUtils reader = null;
CSVUtils maskReader = null;
try {
File file = new File(path);
reader = new CSVUtils(file, separator);
List<String> line;
if (mask != null) {
maskReader = new CSVUtils(new File(mask), separator);
}
List<String> maskLine = null;
while((line = reader.readNext()) != null) {
if (mask != null) {
maskLine = maskReader.readNext();
}
extractObjectFromLine(result, line, 1, maskLine);
}
} catch (FileNotFoundException e) {
e.printStackTrace();
result = null;
}
data = new SimpleData(result);
data.setDataName(name);
MainFrame.getInstance().createDataSession(data);
} catch (Exception e) {
e.printStackTrace();
}
return data;
}
/**
* Construct a DataObject from a set of Strings and add it to a List
* @param list
* the list to add the DataObject to
* @param line
* the line to interpret
*/
private void extractObjectFromLine(List<DataObject> list, List<String> line, int featureNb, List<String> maskLine) {
DataObject obj = new DataObject(1);
double[][] seq = new double[line.size()/featureNb][];
for (int i = 0 ; i < line.size()/featureNb ; i++) {
if (maskLine != null) {
if(Integer.parseInt(maskLine.get(i)) == 0)
continue;
}
seq[i] = new double[featureNb];
for (int j = 0 ; j < featureNb ; j++) {
seq[i][j] = Double.parseDouble(line.get(i*featureNb + j));
}
}
obj.setAttribute(0, new AttributeMultiDimSequence(seq));
list.add(obj);
}
private void extractcustomESV() {
DataCsvExchange.dataToCSV("/home/baptiste/dataExtraite.csv", this.currentDataSession.getData(),
"\t", false);
}
/**
* Return the currently selected session
* @return the DataSession/ImageSession
......
src/mustic/gui/panels/ImageResultPanel.java
View file @ 7482508b
......@@ -1199,12 +1199,6 @@ public class ImageResultPanel extends ResultPanel implements TreeSelectionListen
out.closeEntry();
}
out.putNextEntry(new ZipEntry(name + '/' + "ParameterMode.ser"));
oos = new ObjectOutputStream(out);
oos.writeObject(AttributeMultiDimSequence.getMode());
oos.flush();
out.closeEntry();
out.putNextEntry(new ZipEntry(name + '/' + "ParameterSimplify.ser"));
oos = new ObjectOutputStream(out);
oos.writeObject(AttributeMultiDimSequence.getSimplifyFrom());
......@@ -1311,12 +1305,6 @@ public class ImageResultPanel extends ResultPanel implements TreeSelectionListen
out.closeEntry();
}
out.putNextEntry(new ZipEntry(name + '/' + "ParameterMode.ser"));
oos = new ObjectOutputStream(out);
oos.writeObject(AttributeMultiDimSequence.getMode());
oos.flush();
out.closeEntry();
out.putNextEntry(new ZipEntry(name + '/' + "ParameterSimplify.ser"));
oos = new ObjectOutputStream(out);
oos.writeObject(AttributeMultiDimSequence.getSimplifyFrom());
......
src/mustic/gui/panels/classifier/ClassifierPanel.java
View file @ 7482508b
......@@ -941,20 +941,22 @@ public class ClassifierPanel extends JPanel {
ClassificationWeights weights = new GlobalWeights(data);
if (this.attributeSelectionPanel != null)
weights = this.attributeSelectionPanel.getWeights();
DistanceModel model = null;
/* generates model */
if (this.useNaiveDist) {
// AttributeMultiDimSequence.setMode(AttributeMultiDimSequence.EUCLIDIEN);
model = DistanceModel.generateNaiveModel(data.getOneDataObject());
model = DistanceModel.generateNaiveModel(data.getOneDataObject(), weights);
}
else {
// AttributeMultiDimSequence.setMode(AttributeMultiDimSequence.DTW_BARYCENTRE);
model = DistanceModel.generateDefaultModel(data.getOneDataObject());
model = DistanceModel.generateDefaultModel(data.getOneDataObject(), weights);
}
if (this.attributeSelectionPanel != null)
weights = this.attributeSelectionPanel.getWeights();
int choice = this.modele.getSelectedApproach();
......@@ -1042,7 +1044,6 @@ public class ClassifierPanel extends JPanel {
else {
distanceParameters[th][i] = EmptyDistanceParameter.getInstance();//no parameter for a numerical euclidean distance
averageParameters[i] = EmptyAverageParameter.getInstance();
}
i++;
}
......
src/mustic/gui/panels/classifier/components/HybridClassificationPanel.java
View file @ 7482508b
......@@ -10,7 +10,11 @@ import javax.swing.*;
import javax.swing.border.TitledBorder;
import jcl.data.Data;
import jcl.data.attribute.Attribute;
import jcl.data.attribute.AttributeMultiDimSequence;
import jcl.data.distance.Distance;
import jcl.data.distance.DistanceParameter;
import jcl.data.distance.sequential.DistanceEuclidean;
import jcl.learning.LearningParameters;
import jcl.learning.methods.multistrategy.samarah.*;
import mustic.gui.MainFrame;
......@@ -151,7 +155,11 @@ public class HybridClassificationPanel extends JPanel implements AgentListManage
else {
file += "\t\t<isSimplifySequence>true</isSimplifySequence>\n";
}
if (AttributeMultiDimSequence.getMode() == AttributeMultiDimSequence.EUCLIDIEN) {
for (Distance<Attribute, DistanceParameter> d : agent.getParams().getModel().getDistances()) {
if(!(d instanceof DistanceEuclidean))
}
if (AttributeMultiDimSequence.getMode() == AttributeMultiDimSequence.EUCLIDIEN && agent.getParams().getModel()) {
file += "\t\t<isUseNaiveDist>true</isUseNaiveDist>\n";
}
else {
......
src/mustic/io/ImageData.java
View file @ 7482508b
......@@ -533,12 +533,6 @@ public class ImageData extends SimpleData implements Serializable {
public void setResolution(double resolution) {
this.resolution = resolution;
}
private void setProperties(int mode, int simplifyFrom) {
AttributeMultiDimSequence.setMode(mode);
AttributeMultiDimSequence.setSimplifyFrom(simplifyFrom);
}
/**
* <p>
......
src/mustic/io/ImageMultiDimSequenceData.java deleted 100644 → 0
View file @ 965162f4
package mustic.io;
import java.io.Serializable;
import java.util.ArrayList;
import java.util.Arrays;
import java.util.concurrent.ConcurrentHashMap;
import jcl.data.Data;
import jcl.data.DataObject;
import jcl.data.SimpleData;
import jcl.data.attribute.AttributeMultiDimSequence;
import jcl.data.mask.Mask;
/**
* @author Francois Petitjean
* @deprecated the ImageData class support sequence of images, it should be used instead
*/
public class ImageMultiDimSequenceData extends SimpleData implements Serializable {
private static final long serialVersionUID = 1L;
/**
* position des valeurs dans l'image en cas de masque global
*/
public long[] indexInImage;
public ImageMultiDimSequenceData(RawImage[] seqImages) {
launch(seqImages, null);
}
public ImageMultiDimSequenceData(RawImage[] seqImages, int mode,
int simplifyFrom) {
setProperties(mode, simplifyFrom);
launch(seqImages, null);
}
public ImageMultiDimSequenceData(RawImage[] seqImages, RawImage[] masks) {
launch(seqImages, masks);
}
public ImageMultiDimSequenceData(RawImage[] seqImages, RawImage[] masks,
int mode, int simplifyFrom) {
setProperties(mode, simplifyFrom);
launch(seqImages, masks);
}
private void launch(RawImage[] seqImages, RawImage[] maskImages) {
if (seqImages == null) {
PiLLiTools.error("Invalid image provided", "data/ImageData.java",
"Constructeur aleatoire");
return;
}
if (seqImages[0].getMask() == null) {
if (maskImages == null) {
System.out.println("launchWithoutMasks");
launchWithoutMasks(seqImages);
} else {
System.out.println("launchWithLocalMasksForDTW");
launchWithLocalMasksForDTW(seqImages, maskImages);
}
} else {
if (maskImages == null) {
System.out.println("launchWithGlobalMask");
launchWithGlobalMask(seqImages);
} else {
if (AttributeMultiDimSequence.getMode() == AttributeMultiDimSequence.DTW_BARYCENTRE) {
System.out.println("launchWithGlobalAndLocalMasks");
launchWithGlobalAndLocalMasks(seqImages, maskImages);
} else {
System.err
.println("Demande de masques locaux en euclidien dans "
+ this.getClass().getName()
+ "#launch(RawImage[], RawImage[])");
return;
}
}
}
}
private void launchWithGlobalAndLocalMasks(final RawImage[] seqImages,
final RawImage[] maskImages) {
/* creating the band names */
int nbBands = seqImages[0].getNbSelectedBands();
String[] bandNames = new String[nbBands];
for (int i = 0; i < bandNames.length; i++) {
bandNames[i] = "Band " + i;
}
String[] attNames = { "Sequence" };
this.setAttributesNames(attNames);
Mask globalMask = seqImages[0].getMask();
int nbPixels = seqImages[0].getWidth() * seqImages[0].getHeight()
- seqImages[0].getNbMaskedPixels();
int nbImages = seqImages.length;
System.out.println(nbPixels);
ConcurrentHashMap<Integer, ArrayList<Double[]>> values = new ConcurrentHashMap<Integer, ArrayList<Double[]>>();
/*
* We creates the data array. We need as many DataObject items as pixels
* on image. Each item will contain a tuple.
*/
this.setLocked(new boolean[nbPixels]);
this.currentView = new ArrayList<DataObject>(nbPixels);
this.indexInImage = new long[nbPixels];
int xdim = seqImages[0].getWidth(), ydim = seqImages[0].getHeight();
double[][][] buffer;
double[][][] maskBuffer;
ArrayList<Double[]> tmp;
Double[] tmpTab;
boolean wasNull;
for (int i = 0; i < seqImages.length; i++) {
System.out.println("Image " + i);
buffer = seqImages[i].load(0, 0, seqImages[i].getWidth(),
seqImages[i].getHeight());
seqImages[i] = null;
maskBuffer = maskImages[i].load(0, 0, maskImages[i].getWidth(),
maskImages[i].getHeight());
maskImages[i] = null;
System.gc();
for (int x = 0; x < xdim; x++) {
for (int y = 0; y < ydim; y++) {
if (!globalMask.isMasked(x + y*ydim)) {
if (maskBuffer[0][x][y] == 0) {
tmp = values.get(y * xdim + x);
wasNull = (tmp == null);
if (wasNull) {
tmp = new ArrayList<Double[]>(nbImages / 2);
}
tmpTab = new Double[nbBands];
for (int b = 0; b < nbBands; b++) {
tmpTab[b] = buffer[b][x][y];
}
tmp.add(tmpTab);
if (i == 40)
tmp.trimToSize();
if (wasNull) {
values.put(y * xdim + x, tmp);
}
}