def trainNetwork(self):
self.settings.setValue("ui/ANN/neighborhood", self.spnNeigbourhood.value())
self.settings.setValue("ui/ANN/learningRate", self.spnLearnRate.value())
self.settings.setValue("ui/ANN/maxIterations", self.spnMaxIterations.value())
self.settings.setValue("ui/ANN/topology", self.leTopology.text())
self.settings.setValue("ui/ANN/momentum", self.spnMomentum.value())
self.settings.setValue("ui/ANN/createReport", self.chkCreateReport.isChecked())
self.settings.setValue("ui/ANN/saveSamples", self.chkSaveSamples.isChecked())
self.model = MlpManager(ns=self.spnNeigbourhood.value())
self.model.createMlp(self.inputs["initial"],
self.inputs["factors"].values(),
self.inputs["changeMap"],
[int(n) for n in self.leTopology.text().split(" ")]
)
self.model.setTrainingData(self.inputs["initial"],
self.inputs["factors"].values(),
self.inputs["final"],
mode=self.inputs["samplingMode"],
samples=self.plugin.spnSamplesCount.value()
)
self.model.setEpochs(self.spnMaxIterations.value())
self.model.setValPercent(20)
self.model.setLRate(self.spnLearnRate.value())
self.model.setMomentum(self.spnMomentum.value())
self.model.setContinueTrain()
self.dataTrain = [1]
self.dataVal = [1]
self.plotTrain = self.axes.plot(self.dataTrain,
linewidth=1,
color="green",
)[0]
self.plotVal = self.axes.plot(self.dataVal,
linewidth=1,
color="red",
)[0]
self.model.moveToThread(self.plugin.workThread)
self.plugin.workThread.started.connect(self.model.startTrain)
self.model.updateGraph.connect(self.__updateGraph)
self.model.updateDeltaRMS.connect(self.__updateRMS)
self.model.updateMinValErr.connect(self.__updateValidationError)
self.model.processFinished.connect(self.__trainFinished)
self.model.processFinished.connect(self.plugin.workThread.quit)
self.plugin.workThread.start()
self.inputs["model"] = self.model
def trainNetwork(self):
if not utils.checkInputRasters(self.inputs):
QMessageBox.warning(self.plugin,
self.tr("Missed input data"),
self.tr("Initial or final raster is not set. Please specify input data and try again")
)
return
if not utils.checkFactors(self.inputs):
QMessageBox.warning(self.plugin,
self.tr("Missed input data"),
self.tr("Factors rasters is not set. Please specify them and try again")
)
return
if not utils.checkChangeMap(self.inputs):
QMessageBox.warning(self.plugin,
self.tr("Missed input data"),
self.tr("Change map raster is not set. Please create it try again")
)
return
if self.leTopology.text() == "":
QMessageBox.warning(self.plugin,
self.tr("Wrong network topology"),
self.tr("Network topology is undefined. Please define it and try again")
)
return
self.settings.setValue("ui/ANN/neighborhood", self.spnNeigbourhood.value())
self.settings.setValue("ui/ANN/learningRate", self.spnLearnRate.value())
self.settings.setValue("ui/ANN/maxIterations", self.spnMaxIterations.value())
self.settings.setValue("ui/ANN/topology", self.leTopology.text())
self.settings.setValue("ui/ANN/momentum", self.spnMomentum.value())
self.btnStop.setEnabled(True)
self.btnTrainNetwork.setEnabled(False)
self.plugin.logMessage(self.tr("Init ANN model"))
model = MlpManager(ns=self.spnNeigbourhood.value())
self.inputs["model"] = model
model.createMlp(self.inputs["initial"],
self.inputs["factors"].values(),
self.inputs["changeMap"],
[int(n) for n in self.leTopology.text().split(" ")]
)
self.plugin.logMessage(self.tr("Set training data"))
model.setTrainingData(self.inputs["initial"],
self.inputs["factors"].values(),
self.inputs["changeMap"],
mode=self.inputs["samplingMode"],
samples=self.plugin.spnSamplesCount.value()
)
model.setEpochs(self.spnMaxIterations.value())
model.setValPercent(20)
model.setLRate(self.spnLearnRate.value())
model.setMomentum(self.spnMomentum.value())
model.setContinueTrain()
self.axes.cla()
self.axes.grid(True)
self.dataTrain = []
self.dataVal = []
self.plotTrain = self.axes.plot(self.dataTrain,
linewidth=1,
color="green", marker='o'
)[0]
self.plotVal = self.axes.plot(self.dataVal,
linewidth=1,
color="red",
)[0]
leg = self.axes.legend(('Train', 'Validation'), 'upper right', shadow=False)
for t in leg.get_texts():
t.set_fontsize('small')
model.moveToThread(self.plugin.workThread)
self.plugin.workThread.started.connect(model.startTrain)
self.btnStop.clicked.connect(model.stopTrain)
model.updateGraph.connect(self.__updateGraph)
model.updateDeltaRMS.connect(self.__updateRMS)
model.updateMinValErr.connect(self.__updateValidationError)
model.updateKappa.connect(self.__updateKappa)
model.processInterrupted.connect(self.__trainInterrupted)
model.rangeChanged.connect(self.plugin.setProgressRange)
model.updateProgress.connect(self.plugin.showProgress)
model.errorReport.connect(self.plugin.logErrorReport)
model.processFinished.connect(self.__trainFinished)
model.processFinished.connect(self.plugin.workThread.quit)
self.plugin.logMessage(self.tr("Start trainig ANN model"))
self.plugin.workThread.start()
def trainNetwork(self):
if not utils.checkInputRasters(self.inputs):
QMessageBox.warning(
self.plugin, self.tr("Missed input data"),
self.
tr("Initial or final raster is not set. Please specify input data and try again"
))
return
if not utils.checkFactors(self.inputs):
QMessageBox.warning(
self.plugin, self.tr("Missed input data"),
self.
tr("Factors rasters is not set. Please specify them and try again"
))
return
if not utils.checkChangeMap(self.inputs):
QMessageBox.warning(
self.plugin, self.tr("Missed input data"),
self.
tr("Change map raster is not set. Please create it try again"))
return
if self.leTopology.text() == "":
QMessageBox.warning(
self.plugin, self.tr("Wrong network topology"),
self.
tr("Network topology is undefined. Please define it and try again"
))
return
self.settings.setValue("ui/ANN/neighborhood",
self.spnNeigbourhood.value())
self.settings.setValue("ui/ANN/learningRate",
self.spnLearnRate.value())
self.settings.setValue("ui/ANN/maxIterations",
self.spnMaxIterations.value())
self.settings.setValue("ui/ANN/topology", self.leTopology.text())
self.settings.setValue("ui/ANN/momentum", self.spnMomentum.value())
self.btnStop.setEnabled(True)
self.btnTrainNetwork.setEnabled(False)
self.plugin.logMessage(self.tr("Init ANN model"))
model = MlpManager(ns=self.spnNeigbourhood.value())
self.inputs["model"] = model
model.createMlp(self.inputs["initial"],
self.inputs["factors"].values(),
self.inputs["changeMap"],
[int(n) for n in self.leTopology.text().split(" ")])
self.plugin.logMessage(self.tr("Set training data"))
model.setTrainingData(self.inputs["initial"],
self.inputs["factors"].values(),
self.inputs["changeMap"],
mode=self.inputs["samplingMode"],
samples=self.plugin.spnSamplesCount.value())
model.setEpochs(self.spnMaxIterations.value())
model.setValPercent(20)
model.setLRate(self.spnLearnRate.value())
model.setMomentum(self.spnMomentum.value())
model.setContinueTrain()
self.axes.cla()
self.axes.grid(True)
self.dataTrain = []
self.dataVal = []
self.plotTrain = self.axes.plot(self.dataTrain,
linewidth=1,
color="green",
marker='o')[0]
self.plotVal = self.axes.plot(
self.dataVal,
linewidth=1,
color="red",
)[0]
leg = self.axes.legend(('Train', 'Validation'),
'upper right',
shadow=False)
for t in leg.get_texts():
t.set_fontsize('small')
model.moveToThread(self.plugin.workThread)
self.plugin.workThread.started.connect(model.startTrain)
self.btnStop.clicked.connect(model.stopTrain)
model.updateGraph.connect(self.__updateGraph)
model.updateDeltaRMS.connect(self.__updateRMS)
model.updateMinValErr.connect(self.__updateValidationError)
model.updateKappa.connect(self.__updateKappa)
model.processInterrupted.connect(self.__trainInterrupted)
model.rangeChanged.connect(self.plugin.setProgressRange)
model.updateProgress.connect(self.plugin.showProgress)
model.errorReport.connect(self.plugin.logErrorReport)
model.processFinished.connect(self.__trainFinished)
model.processFinished.connect(self.plugin.workThread.quit)
self.plugin.logMessage(self.tr("Start trainig ANN model"))
self.plugin.workThread.start()
class NeuralNetworkWidget(QWidget, Ui_Widget):
def __init__(self, plugin, parent=None):
QWidget.__init__(self, parent)
self.setupUi(self)
self.plugin = plugin
self.inputs = plugin.inputs
self.settings = QSettings("NextGIS", "MOLUSCE")
# init plot for learning curve
self.figure = Figure()
self.axes = self.figure.add_subplot(111)
self.axes.grid(True)
self.figure.suptitle(self.tr("Neural Network learning curve"))
self.canvas = FigureCanvas(self.figure)
self.mpltoolbar = NavigationToolbar(self.canvas, None)
lstActions = self.mpltoolbar.actions()
self.mpltoolbar.removeAction(lstActions[7])
self.layoutPlot.addWidget(self.canvas)
self.layoutPlot.addWidget(self.mpltoolbar)
# and configure matplotlib params
rcParams['font.serif'] = "Verdana, Arial, Liberation Serif"
rcParams['font.sans-serif'] = "Tahoma, Arial, Liberation Sans"
rcParams['font.cursive'] = "Courier New, Arial, Liberation Sans"
rcParams['font.fantasy'] = "Comic Sans MS, Arial, Liberation Sans"
rcParams['font.monospace'] = "Courier New, Liberation Mono"
self.chkCreateReport.toggled.connect(self.__toggleLineEdit)
self.chkSaveSamples.toggled.connect(self.__toggleLineEdit)
self.btnSelectReport.clicked.connect(self.__selectFile)
self.btnSelectSamples.clicked.connect(self.__selectFile)
self.btnTrainNetwork.clicked.connect(self.trainNetwork)
self.manageGui()
def manageGui(self):
self.spnNeigbourhood.setValue(self.settings.value("ui/ANN/neighborhood", 1).toInt()[0])
self.spnLearnRate.setValue(self.settings.value("ui/ANN/learningRate", 0.1).toFloat()[0])
self.spnMaxIterations.setValue(self.settings.value("ui/ANN/maxIterations", 1000).toInt()[0])
self.leTopology.setText(self.settings.value("ui/ANN/topology", "10").toString())
self.spnMomentum.setValue(self.settings.value("ui/ANN/momentum", 0.05).toFloat()[0])
self.chkCreateReport.setChecked(self.settings.value("ui/ANN/createReport", False).toBool())
self.chkSaveSamples.setChecked(self.settings.value("ui/ANN/saveSamples", False).toBool())
def trainNetwork(self):
self.settings.setValue("ui/ANN/neighborhood", self.spnNeigbourhood.value())
self.settings.setValue("ui/ANN/learningRate", self.spnLearnRate.value())
self.settings.setValue("ui/ANN/maxIterations", self.spnMaxIterations.value())
self.settings.setValue("ui/ANN/topology", self.leTopology.text())
self.settings.setValue("ui/ANN/momentum", self.spnMomentum.value())
self.settings.setValue("ui/ANN/createReport", self.chkCreateReport.isChecked())
self.settings.setValue("ui/ANN/saveSamples", self.chkSaveSamples.isChecked())
self.model = MlpManager(ns=self.spnNeigbourhood.value())
self.model.createMlp(self.inputs["initial"],
self.inputs["factors"].values(),
self.inputs["changeMap"],
[int(n) for n in self.leTopology.text().split(" ")]
)
self.model.setTrainingData(self.inputs["initial"],
self.inputs["factors"].values(),
self.inputs["final"],
mode=self.inputs["samplingMode"],
samples=self.plugin.spnSamplesCount.value()
)
self.model.setEpochs(self.spnMaxIterations.value())
self.model.setValPercent(20)
self.model.setLRate(self.spnLearnRate.value())
self.model.setMomentum(self.spnMomentum.value())
self.model.setContinueTrain()
self.dataTrain = [1]
self.dataVal = [1]
self.plotTrain = self.axes.plot(self.dataTrain,
linewidth=1,
color="green",
)[0]
self.plotVal = self.axes.plot(self.dataVal,
linewidth=1,
color="red",
)[0]
self.model.moveToThread(self.plugin.workThread)
self.plugin.workThread.started.connect(self.model.startTrain)
self.model.updateGraph.connect(self.__updateGraph)
self.model.updateDeltaRMS.connect(self.__updateRMS)
self.model.updateMinValErr.connect(self.__updateValidationError)
self.model.processFinished.connect(self.__trainFinished)
self.model.processFinished.connect(self.plugin.workThread.quit)
self.plugin.workThread.start()
self.inputs["model"] = self.model
def __trainFinished(self):
self.plugin.workThread.started.disconnect(self.model.startTrain)
def __updateRMS(self, dRMS):
self.leDeltaRMS.setText(QString.number(dRMS))
def __updateValidationError(self, error):
self.leValidationError.setText(QString.number(error))
def __updateGraph(self, errTrain, errVal):
self.dataTrain.append(errTrain)
self.dataVal.append(errVal)
ymin = min([min(self.dataTrain), min(self.dataVal)])
ymax = max([max(self.dataTrain), max(self.dataVal)])
self.axes.set_xbound(lower=0, upper=len(self.dataVal))
self.axes.set_ybound(lower=ymin, upper=ymax)
self.plotTrain.set_xdata(numpy.arange(len(self.dataTrain)))
self.plotTrain.set_ydata(numpy.array(self.dataTrain))
self.plotVal.set_xdata(numpy.arange(len(self.dataVal)))
self.plotVal.set_ydata(numpy.array(self.dataVal))
self.canvas.draw()
def __selectFile(self):
senderName = self.sender().objectName()
# TODO: implement dialog for necessary data type
fileName = utils.saveRasterDialog(self,
self.settings,
self.tr("Save file"),
self.tr("GeoTIFF (*.tif *.tiff *.TIF *.TIFF)")
)
if fileName.isEmpty():
return
if senderName == "btnSelectReport":
self.leReportPath.setText(fileName)
elif senderName == "btnSelectSamples":
self.leSamplesPath.setText(fileName)
def __toggleLineEdit(self, checked):
senderName = self.sender().objectName()
if senderName == "chkCreateReport":
if checked:
self.leReportPath.setEnabled(True)
self.btnSelectReport.setEnabled(True)
else:
self.leReportPath.setEnabled(False)
self.btnSelectReport.setEnabled(False)
elif senderName == "chkSaveSamples":
if checked:
self.leSamplesPath.setEnabled(True)
self.btnSelectSamples.setEnabled(True)
else:
self.leSamplesPath.setEnabled(False)
self.btnSelectSamples.setEnabled(False)
