saberbud/q3DMASC
1
0
Fork 0
mirror of https://github.com/dgirardeau/q3DMASC.git synced 2026-08-29 16:40:49 +08:00
Code Issues Packages Projects Releases Wiki Activity
Files
8ec8850590042f514383283ba259b226b73ef960
q3DMASC/q3DMASCTools.cpp
T
Daniel Girardeau-Montaut 8ec8850590 Bug fixes
2019-03-28 21:29:03 +01:00

1373 lines
38 KiB
C++
Raw Blame History

//##########################################################################
//# #
//# CLOUDCOMPARE PLUGIN: q3DMASC #
//# #
//# This program is free software; you can redistribute it and/or modify #
//# it under the terms of the GNU General Public License as published by #
//# the Free Software Foundation; version 2 or later of the License. #
//# #
//# This program is distributed in the hope that it will be useful, #
//# but WITHOUT ANY WARRANTY; without even the implied warranty of #
//# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the #
//# GNU General Public License for more details. #
//# #
//# COPYRIGHT: Dimitri Lague / CNRS / UEB #
//# #
//##########################################################################
#include "q3DMASCTools.h"
//Local
#include "PointFeature.h"
#include "NeighborhoodFeature.h"
#include "DualCloudFeature.h"
#include "ContextBasedFeature.h"
//qCC_io
#include <FileIOFilter.h>
//qCC_db
#include <ccScalarField.h>
#include <ccPointCloud.h>
//Qt
#include <QTextStream>
#include <QFile>
#include <QFileInfo>
#include <QDir>
#include <QMutex>
//system
#include <assert.h>
using namespace masc;
bool Tools::SaveClassifier( QString filename,
const Feature::Set& features,
const QString corePointsRole,
const masc::Classifier& classifier,
QWidget* parent/*=nullptr*/)
{
//first save the classifier data (same base filename but with the ymal extension)
QFileInfo fi(filename);
QString yamlFilename = fi.baseName() + ".yaml";
QString yamlAbsoluteFilename = fi.absoluteDir().absoluteFilePath(yamlFilename);
if (!classifier.toFile(yamlAbsoluteFilename, parent))
{
ccLog::Error("Failed to save the classifier data");
return false;
}
QFile file(filename);
if (!file.open(QFile::Text | QFile::WriteOnly))
{
ccLog::Warning(QString("Can't open file '%1' for writing").arg(filename));
return false;
}
QTextStream stream(&file);
stream << "# 3DMASC classifier file" << endl;
stream << "classifier: " << yamlFilename << endl;
//look for all clouds (labels)
QList<QString> cloudLabels;
for (Feature::Shared f : features)
{
if (f->cloud1 && !cloudLabels.contains(f->cloud1Label))
cloudLabels.push_back(f->cloud1Label);
if (f->cloud2 && !cloudLabels.contains(f->cloud2Label))
cloudLabels.push_back(f->cloud2Label);
}
if (!corePointsRole.isEmpty() && !cloudLabels.contains(corePointsRole))
{
cloudLabels.push_back(corePointsRole);
}
stream << "# Clouds (roles)" << endl;
for (const QString& label : cloudLabels)
{
stream << "cloud: " << label << endl;
}
if (!corePointsRole.isEmpty())
{
stream << "# Core points (classified role)" << endl;
stream << "core_points: " << corePointsRole << endl;
}
stream << "# Features" << endl;
for (Feature::Shared f : features)
{
stream << "feature: " << f->toString() << endl;
}
return true;
}
bool Tools::LoadClassifierCloudLabels(QString filename, QList<QString>& labels, QString& corePointsLabel, bool& filenamesSpecified)
{
//just in case
corePointsLabel.clear();
labels.clear();
QFile file(filename);
if (!file.open(QFile::Text | QFile::ReadOnly))
{
ccLog::Warning(QString("Can't open file '%1'").arg(filename));
return false;
}
QTextStream stream(&file);
int filenameCount = 0;
for (int lineNumber = 0; ; ++lineNumber)
{
QString line = stream.readLine();
if (line.isNull())
{
//eof
break;
}
++lineNumber;
line = line.toUpper();
if (line.startsWith("CLOUD:"))
{
QString command = line.mid(6).trimmed();
QStringList tokens = command.split('=');
if (tokens.size() == 0)
{
ccLog::Warning("Malformed file: expecting some tokens after 'cloud:' on line #" + QString::number(lineNumber));
return false;
}
QString label = tokens.front();
if (labels.contains(label))
{
ccLog::Warning(QString("Malformed file: role '%1:' is already defined/used on line #%2").arg(label).arg(lineNumber));
return false;
}
labels.push_back(label);
if (tokens.size() > 1)
++filenameCount;
}
else if (line.startsWith("CORE_POINTS:"))
{
if (!corePointsLabel.isEmpty())
{
//core points defined multiple times?!
continue;
}
QString command = line.mid(12);
QStringList tokens = command.split('_');
if (tokens.empty())
{
ccLog::Warning("Malformed file: expecting tokens after 'core_points:' on line #" + QString::number(lineNumber));
return false;
}
corePointsLabel = tokens[0].trimmed();
}
}
filenamesSpecified = (filenameCount > 0 && filenameCount == labels.size());
return true;
}
static bool CreateFeaturesFromCommand(const QString& command, QString corePointsRole, int lineNumber, const Tools::NamedClouds& clouds, std::vector<Feature::Shared>& rawFeatures, std::vector<double>& scales)
{
QStringList tokens = command.split('_');
if (tokens.empty())
{
ccLog::Warning("Malformed file: expecting at least one token after 'feature:' on line #" + QString::number(lineNumber));
return false;
}
Feature::Shared feature;
//read the type
QString typeStr = tokens[0].trimmed().toUpper();
{
for (int iteration = 0; iteration < 1; ++iteration) //fake loop for easy break
{
PointFeature::PointFeatureType pointFeatureType = PointFeature::FromUpperString(typeStr);
if (pointFeatureType != PointFeature::Invalid)
{
//we have a point feature
PointFeature* pointFeature = new PointFeature(pointFeatureType);
//specific case: 'SF'
if (pointFeatureType == PointFeature::SF)
{
QString sfIndexStr = typeStr.mid(2);
bool ok = true;
int sfIndex = sfIndexStr.toInt(&ok);
if (!ok)
{
ccLog::Warning(QString("Malformed file: expecting a valid integer value after 'SF' on line #%1").arg(lineNumber));
delete pointFeature;
return false;
}
pointFeature->sourceSFIndex = sfIndex;
}
feature.reset(pointFeature);
break;
}
NeighborhoodFeature::NeighborhoodFeatureType neighborhoodFeatureType = NeighborhoodFeature::FromUpperString(typeStr);
if (neighborhoodFeatureType != NeighborhoodFeature::Invalid)
{
//we have a neighborhood feature
feature = NeighborhoodFeature::Shared(new NeighborhoodFeature(neighborhoodFeatureType));
break;
}
ContextBasedFeature::ContextBasedFeatureType contextBasedFeatureType = ContextBasedFeature::FromUpperString(typeStr);
if (contextBasedFeatureType != ContextBasedFeature::Invalid)
{
//we have a context-based feature
QString featureStr = ContextBasedFeature::ToString(contextBasedFeatureType);
int kNN = 1;
if (featureStr.length() < typeStr.length())
{
bool ok = false;
kNN = typeStr.mid(featureStr.length()).toInt(&ok);
if (!ok || kNN <= 0)
{
ccLog::Warning(QString("Malformed file: expecting a valid and positive number after '%1' on line #%2").arg(featureStr).arg(lineNumber));
return false;
}
}
feature = ContextBasedFeature::Shared(new ContextBasedFeature(contextBasedFeatureType, kNN));
break;
}
DualCloudFeature::DualCloudFeatureType dualCloudFeatureType = DualCloudFeature::FromUpperString(typeStr);
if (dualCloudFeatureType != DualCloudFeature::Invalid)
{
//we have a dual cloud feature
feature = DualCloudFeature::Shared(new DualCloudFeature(dualCloudFeatureType));
break;
}
if (!feature)
{
ccLog::Warning(QString("Malformed file: unrecognized token '%1' after 'feature:' on line #%2").arg(typeStr).arg(lineNumber));
return false;
}
}
}
assert(feature);
//read the scales
bool useAllScales = false;
{
QString scaleStr = tokens[1].toUpper();
if (!scaleStr.startsWith("SC"))
{
ccLog::Warning(QString("Malformed file: unrecognized token '%1' (expecting the scale descriptor 'SC...' on line #%2").arg(typeStr).arg(lineNumber));
return false;
}
if (scaleStr == "SC0")
{
//no scale
}
else if (scaleStr == "SCX")
{
//all scales
useAllScales = true;
}
else
{
//read the specific scale index
bool ok = true;
feature->scale = scaleStr.mid(2).toDouble(&ok);
if (!ok)
{
ccLog::Warning(QString("Malformed file: expecting a valid number after 'SC:' on line #%1").arg(lineNumber));
return false;
}
}
}
//process the next tokens (may not be ordered)
int cloudCount = 0;
bool statDefined = false;
bool mathDefined = false;
for (int i = 2; i < tokens.size(); ++i)
{
QString token = tokens[i].trimmed().toUpper();
//is the token a 'stat' one?
if (!statDefined)
{
if (token == "MEAN")
{
feature->stat = Feature::MEAN;
statDefined = true;
}
else if (token == "MODE")
{
feature->stat = Feature::MODE;
statDefined = true;
}
else if (token == "MEDIAN")
{
feature->stat = Feature::MEDIAN;
statDefined = true;
}
else if (token == "STD")
{
feature->stat = Feature::STD;
statDefined = true;
}
else if (token == "RANGE")
{
feature->stat = Feature::RANGE;
statDefined = true;
}
else if (token == "SKEW")
{
feature->stat = Feature::SKEW;
statDefined = true;
}
if (statDefined)
{
continue;
}
}
//is the token a cloud name?
if (cloudCount < 2)
{
bool cloudNameMatches = false;
for (QMap<QString, ccPointCloud* >::const_iterator it = clouds.begin(); it != clouds.end(); ++it)
{
QString key = it.key().toUpper();
if (key == token)
{
if (cloudCount == 0)
{
feature->cloud1 = it.value();
feature->cloud1Label = key;
}
else if (cloudCount == 1)
{
feature->cloud2 = it.value();
feature->cloud2Label = key;
if (feature && feature->getType() == Feature::Type::ContextBasedFeature)
{
//context cloud is always the second one for context based features
//and the class is just after the cloud name
if (i + 1 < tokens.size())
{
bool ok = false;
int classLabel = tokens[i + 1].toInt(&ok);
if (!ok)
{
ccLog::Warning(QString("Malformed file: expecting a class number after the context cloud '%1' on line #%2").arg(token).arg(lineNumber));
return false;
}
qSharedPointerCast<ContextBasedFeature>(feature)->ctxClassLabel = classLabel;
++i;
}
}
}
else
{
//we can't fall here
assert(false);
}
++cloudCount;
cloudNameMatches = true;
break;
}
}
if (cloudNameMatches)
{
continue;
}
}
//is the token a 'math' one?
if (!mathDefined)
{
if (token == "MINUS")
{
feature->op = Feature::MINUS;
mathDefined = true;
}
else if (token == "PLUS")
{
feature->op = Feature::PLUS;
mathDefined = true;
}
else if (token == "DIVIDE")
{
feature->op = Feature::DIVIDE;
mathDefined = true;
}
else if (token == "MULTIPLY")
{
feature->op = Feature::MULTIPLY;
mathDefined = true;
}
if (mathDefined)
{
continue;
}
}
//is the token a 'context' descriptor?
if (feature->getType() == Feature::Type::ContextBasedFeature && token.startsWith("CTX"))
{
//read the context label
QString ctxLabelStr = token.mid(2);
bool ok = true;
int ctxLabel = ctxLabelStr.toInt(&ok);
if (!ok)
{
ccLog::Warning(QString("Malformed file: expecting a valid integer value after 'CTX' on line #%1").arg(lineNumber));
return false;
}
static_cast<ContextBasedFeature*>(feature.data())->ctxClassLabel = ctxLabel;
continue;
}
//if we are here, it means we couldn't find a correspondance for the current token
ccLog::Warning(QString("Malformed file: unrecognized or unexpected token '%1' on line #%2").arg(token).arg(lineNumber));
return false;
}
//now create the various versions of rules (if any)
if (useAllScales)
{
if (scales.empty())
{
ccLog::Warning("Malformed file: 'SCx' token used while no scale is defined" + QString(" (line %1)").arg(lineNumber));
return false;
}
feature->scale = scales.front();
//we will duplicate the original feature AFTER having checked its consistency!
}
//now check the consistency of the rule
QString errorMessage;
if (!feature->checkValidity(corePointsRole, errorMessage))
{
ccLog::Warning("Malformed feature: " + errorMessage + QString(" (line %1)").arg(lineNumber));
return false;
}
//save it
rawFeatures.push_back(feature);
if (useAllScales)
{
for (size_t i = 1; i < scales.size(); ++i)
{
//copy the original rule
Feature::Shared newFeature = feature->clone();
newFeature->scale = scales.at(i);
//as we only change the scale value, all the duplicated features should be valid
assert(newFeature->checkValidity(corePointsRole, errorMessage));
rawFeatures.push_back(newFeature);
}
}
return true;
}
static bool ReadScales(const QString& command, std::vector<double>& scales, int lineNumber)
{
assert(scales.empty());
QStringList tokens = command.split(';');
if (tokens.empty())
{
ccLog::Warning("Malformed file: expecting at least one token after 'scales:' on line #" + QString::number(lineNumber));
return false;
}
for (const QString& token : tokens)
{
if (token.contains(':'))
{
//it's probably a range
QStringList subTokens = token.trimmed().split(':');
if (subTokens.size() != 3)
{
ccLog::Warning(QString("Malformed file: expecting 3 tokens for a range of scales (%1)").arg(token));
return false;
}
bool ok[3] = { true, true, true };
double start = subTokens[0].trimmed().toDouble(ok);
double step = subTokens[1].toDouble(ok + 1);
double stop = subTokens[2].toDouble(ok + 2);
if (!ok[0] || !ok[1] || !ok[2])
{
ccLog::Warning(QString("Malformed file: invalid values in scales range (%1) on line #%2").arg(token).arg(lineNumber));
return false;
}
if (stop < start || step <= 1.0 - 6)
{
ccLog::Warning(QString("Malformed file: invalid range (%1) on line #%2").arg(token).arg(lineNumber));
return false;
}
for (double v = start; v <= stop + 1.0e-6; v += step)
{
scales.push_back(v);
}
}
else
{
bool ok = true;
double v = token.trimmed().toDouble(&ok);
if (!ok)
{
ccLog::Warning(QString("Malformed file: invalid scale value (%1) on line #%2").arg(token).arg(lineNumber));
return false;
}
scales.push_back(v);
}
}
scales.shrink_to_fit();
return true;
}
static bool ReadCorePoints(const QString& command, const Tools::NamedClouds& clouds, masc::CorePoints& corePoints, int lineNumber)
{
QStringList tokens = command.split('_');
if (tokens.empty())
{
ccLog::Warning("Malformed file: expecting tokens after 'core_points:' on line #" + QString::number(lineNumber));
return false;
}
QString pcName = tokens[0].trimmed();
if (!clouds.contains(pcName))
{
ccLog::Warning(QString("Malformed file: unknown cloud '%1' on line #%2 (make sure it is declared before the core points)").arg(pcName).arg(lineNumber));
return false;
}
corePoints.origin = clouds[pcName];
corePoints.role = pcName;
//should we sub-sample the origin cloud?
if (tokens.size() > 1)
{
if (tokens[1].toUpper() == "SS")
{
if (tokens.size() < 3)
{
ccLog::Warning("Malformed file: missing token after 'SS' on line #" + QString::number(lineNumber));
return false;
}
QString options = tokens[2];
if (options.startsWith('R'))
{
corePoints.selectionMethod = CorePoints::RANDOM;
}
else if (options.startsWith('S'))
{
corePoints.selectionMethod = CorePoints::SPATIAL;
}
else
{
ccLog::Warning("Malformed file: unknown option after 'SS' on line #" + QString::number(lineNumber));
return false;
}
//read the subsampling parameter (ignore the first character)
bool ok = false;
corePoints.selectionParam = options.mid(1).toDouble(&ok);
if (!ok)
{
ccLog::Warning("Malformed file: expecting a number after 'SS_X' on line #" + QString::number(lineNumber));
return false;
}
} //end of subsampling options
}
return true;
}
static bool ReadCloud(const QString& command, Tools::NamedClouds& clouds, QDir& defaultDir, int lineNumber, FileIOFilter::LoadParameters& loadParameters)
{
QStringList tokens = command.split('=');
if (tokens.size() != 2)
{
ccLog::Warning("Malformed file: expecting 2 tokens after 'cloud:' on line #" + QString::number(lineNumber));
return false;
}
QString pcName = tokens[0].trimmed();
QString pcFilename = defaultDir.absoluteFilePath(tokens[1].trimmed());
//try to open the cloud
{
CC_FILE_ERROR error = CC_FERR_NO_ERROR;
ccHObject* object = FileIOFilter::LoadFromFile(pcFilename, loadParameters, error);
if (error != CC_FERR_NO_ERROR || !object)
{
//error message already issued
if (object)
delete object;
return false;
}
ccHObject::Container cloudsInFile;
object->filterChildren(cloudsInFile, false, CC_TYPES::POINT_CLOUD, true);
if (cloudsInFile.empty())
{
ccLog::Warning("File doesn't contain a single cloud");
delete object;
return false;
}
else if (cloudsInFile.size() > 1)
{
ccLog::Warning("File contains more than one cloud, only the first one will be kept");
}
ccPointCloud* pc = static_cast<ccPointCloud*>(cloudsInFile.front());
for (size_t i = 1; i < cloudsInFile.size(); ++i)
{
delete cloudsInFile[i];
}
if (pc->getParent())
pc->getParent()->detachChild(pc);
pc->setName(pcName); //DGM: warning, may not be acceptable in the GUI version?
clouds.insert(pcName, pc);
}
return true;
}
bool Tools::LoadFile( const QString& filename,
Tools::NamedClouds* clouds,
bool cloudsAreProvided,
std::vector<Feature::Shared>* rawFeatures/*=nullptr*/, //requires 'clouds'
masc::CorePoints* corePoints/*=nullptr*/, //requires 'clouds'
masc::Classifier* classifier/*=nullptr*/,
TrainParameters* parameters/*=nullptr*/,
QWidget* parent/*=nullptr*/)
{
QFileInfo fi(filename);
if (!fi.exists())
{
ccLog::Warning(QString("Can't find file '%1'").arg(filename));
return false;
}
QFile file(filename);
if (!file.open(QFile::Text | QFile::ReadOnly))
{
ccLog::Warning(QString("Can't open file '%1'").arg(filename));
return false;
}
//to use the same 'global shift' for multiple files
CCVector3d loadCoordinatesShift(0, 0, 0);
bool loadCoordinatesTransEnabled = false;
FileIOFilter::LoadParameters loadParameters;
if (!cloudsAreProvided)
{
loadParameters.alwaysDisplayLoadDialog = true;
loadParameters.shiftHandlingMode = ccGlobalShiftManager::DIALOG_IF_NECESSARY;
loadParameters.coordinatesShift = &loadCoordinatesShift;
loadParameters.coordinatesShiftEnabled = &loadCoordinatesTransEnabled;
loadParameters.parentWidget = parent;
FileIOFilter::ResetSesionCounter();
}
try
{
assert(!rawFeatures || rawFeatures->empty());
std::vector<double> scales;
QTextStream stream(&file);
bool badFeatures = false;
for (int lineNumber = 1; ; ++lineNumber)
{
QString line = stream.readLine();
if (line.isNull())
{
//eof
break;
}
if (line.startsWith("#"))
{
//comment
continue;
}
//strip out the potential comment at the end of the line as well
int commentIndex = line.indexOf('#');
if (commentIndex >= 0)
line = line.left(commentIndex);
QString upperLine = line.toUpper();
if (upperLine.startsWith("CLASSIFIER:")) //classifier
{
if (!classifier)
{
//no need to load the classifier
continue;
}
if (classifier->isValid())
{
ccLog::Warning("Malformed file: can't declare the classifier file twice! (line #" + QString::number(lineNumber) + ")");
return false;
}
QString yamlFilename = line.mid(11).trimmed();
QString yamlAbsoluteFilename = fi.absoluteDir().absoluteFilePath(yamlFilename);
if (!classifier->fromFile(yamlAbsoluteFilename, parent))
{
ccLog::Warning("Failed to load the classifier file from " + yamlAbsoluteFilename);
return false;
}
ccLog::Print("[3DMASC] Classifier data loaded from " + yamlAbsoluteFilename);
}
else if (upperLine.startsWith("CLOUD:")) //clouds
{
if (!clouds || cloudsAreProvided)
{
//no need to load the clouds in this case
continue;
}
QString command = line.mid(6);
if (!ReadCloud(command, *clouds, fi.absoluteDir(), lineNumber, loadParameters))
{
return false;
}
}
else if (upperLine.startsWith("TEST:")) //test cloud
{
if (!clouds || cloudsAreProvided)
{
//no need to load the clouds in this case
continue;
}
QString command = line.mid(5);
if (!ReadCloud("TEST=" + command, *clouds, fi.absoluteDir(), lineNumber, loadParameters)) //add the TEST keyword so that the cloud will be loaded as the TEST cloud
{
return false;
}
}
else if (upperLine.startsWith("CORE_POINTS:")) //core points
{
if (!corePoints)
{
//no need to load the core points
continue;
}
if (corePoints->origin)
{
ccLog::Warning("Core points already defined (those declared on line #" + QString::number(lineNumber) + " will be ignored)");
}
else
{
QString command = line.mid(12);
if (clouds && !ReadCorePoints(command, *clouds, *corePoints, lineNumber))
{
return false;
}
}
}
else if (upperLine.startsWith("SCALES:")) //scales
{
if (!scales.empty())
{
ccLog::Warning("Malformed file: scales defined twice (line #" + QString::number(lineNumber) + ")");
return false;
}
QString command = line.mid(7);
if (!ReadScales(command, scales, lineNumber))
{
return false;
}
}
else if (upperLine.startsWith("FEATURE:")) //feature
{
QString command = line.mid(8);
if (rawFeatures && clouds)
{
if (!CreateFeaturesFromCommand(command, corePoints ? corePoints->role : QString(), lineNumber, *clouds, *rawFeatures, scales))
{
//error message already issued
//return false;
badFeatures = true; //we continue as we want to get ALL the errors
}
}
}
else if (upperLine.startsWith("PARAM_")) //parameter
{
if (parameters) //no need to actually read the parameters if the caller didn't requested them
{
QStringList tokens = upperLine.split("=");
if (tokens.size() != 2)
{
ccLog::Warning(QString("Line #%1: malformed parameter command (expecting param_XXX=Y)").arg(lineNumber));
return false;
}
bool ok = false;
if (tokens[0] == "PARAM_MAX_DEPTH")
{
parameters->rt.maxDepth = tokens[1].toInt(&ok);
}
else if (tokens[0] == "PARAM_MAX_TREE_COUNT")
{
parameters->rt.maxTreeCount = tokens[1].toInt(&ok);
}
else if (tokens[0] == "PARAM_ACTIVE_VAR_COUNT")
{
parameters->rt.activeVarCount = tokens[1].toInt(&ok);
}
else if (tokens[0] == "PARAM_MIN_SAMPLE_COUNT")
{
parameters->rt.minSampleCount = tokens[1].toInt(&ok);
}
else if (tokens[0] == "PARAM_TEST_DATA_RATIO")
{
parameters->testDataRatio = tokens[1].toFloat(&ok);
}
else
{
ccLog::Warning(QString("Line #%1: unrecognized parameter: ").arg(lineNumber) + tokens[0]);
}
if (!ok)
{
ccLog::Warning(QString("Line #%1: invalid value for parameter ").arg(lineNumber) + tokens[0]);
}
}
}
else
{
ccLog::Warning(QString("Line #%1: unrecognized token/command: ").arg(lineNumber) + (line.length() < 10 ? line : line.left(10) + "..."));
return false;
}
}
if (badFeatures)
{
return false;
}
}
catch (const std::bad_alloc&)
{
ccLog::Warning("Not enough memory");
return false;
}
if (rawFeatures)
rawFeatures->shrink_to_fit();
return true;
}
bool Tools::LoadClassifier(QString filename, NamedClouds& clouds, Feature::Set& rawFeatures, masc::Classifier& classifier, QWidget* parent/*=nullptr*/)
{
return LoadFile(filename, &clouds, true, &rawFeatures, nullptr, &classifier, nullptr, parent);
}
bool Tools::LoadTrainingFile( QString filename,
Feature::Set& rawFeatures,
NamedClouds& loadedClouds,
TrainParameters& parameters,
CorePoints* corePoints/*=nullptr*/,
QWidget* parentWidget/*=nullptr*/)
{
bool cloudsWereProvided = !loadedClouds.empty();
if (LoadFile(filename, &loadedClouds, cloudsWereProvided, &rawFeatures, corePoints, nullptr, &parameters, parentWidget))
{
return true;
}
else
{
if (!cloudsWereProvided)
{
//delete the already loaded clouds (if any)
for (NamedClouds::const_iterator it = loadedClouds.begin(); it != loadedClouds.end(); ++it)
delete it.value();
}
return false;
}
}
CCLib::ScalarField* Tools::RetrieveSF(const ccPointCloud* cloud, const QString& sfName, bool caseSensitive/*=true*/)
{
if (!cloud)
{
assert(false);
return nullptr;
}
int sfIdx = -1;
if (caseSensitive)
{
sfIdx = cloud->getScalarFieldIndexByName(qPrintable(sfName));
}
else
{
QString sfNameUpper = sfName.toUpper();
for (unsigned i = 0; i < cloud->getNumberOfScalarFields(); ++i)
{
if (QString(cloud->getScalarField(i)->getName()).toUpper() == sfNameUpper)
{
sfIdx = static_cast<int>(i);
break;
}
}
}
if (sfIdx >= 0)
{
return cloud->getScalarField(sfIdx);
}
else
{
return nullptr;
}
}
struct FeaturesAndScales
{
std::vector<double> scales;
size_t featureCount = 0;
QMap<double, std::vector<PointFeature::Shared> > pointFeaturesPerScale;
QMap<double, std::vector<NeighborhoodFeature::Shared> > neighborhoodFeaturesPerScale;
QMap<double, std::vector<ContextBasedFeature::Shared> > contextBasedFeaturesPerScale;
};
bool Tools::PrepareFeatures(const CorePoints& corePoints, Feature::Set& features, QString& errorStr, CCLib::GenericProgressCallback* progressCb/*=nullptr*/, SFCollector* generatedScalarFields/*=nullptr*/)
{
if (features.empty() || !corePoints.origin)
{
//invalid input parameters
assert(false);
return false;
}
//gather all the scales that need to be extracted
QMap<ccPointCloud*, FeaturesAndScales> cloudsWithScaledFeatures;
//and prepare the features (scalar fields, etc.) at the same time
for (const Feature::Shared& feature : features)
{
QString errorMessage("invalid pointer");
assert(!corePoints.role.isEmpty());
if (!feature || !feature->checkValidity(corePoints.role, errorMessage))
{
errorStr = "Invalid rule/feature: " + errorMessage;
return false;
}
//prepare the feature
if (!feature->prepare(corePoints, errorStr, progressCb, generatedScalarFields))
{
//something failed (error should be up to date)
return false;
}
if (feature->scaled())
{
try
{
switch (feature->getType())
{
//Point features
case Feature::Type::PointFeature:
{
//build the scaled feature list attached to the first cloud
if (feature->cloud1)
{
FeaturesAndScales& fas = cloudsWithScaledFeatures[feature->cloud1];
fas.pointFeaturesPerScale[feature->scale].push_back(qSharedPointerCast<PointFeature>(feature));
++fas.featureCount;
if (std::find(fas.scales.begin(), fas.scales.end(), feature->scale) == fas.scales.end())
{
fas.scales.push_back(feature->scale);
}
}
//build the scaled feature list attached to the second cloud (if any)
if (feature->cloud2 && feature->cloud2 != feature->cloud1 && feature->op != Feature::NO_OPERATION)
{
FeaturesAndScales& fas = cloudsWithScaledFeatures[feature->cloud2];
++fas.featureCount;
fas.pointFeaturesPerScale[feature->scale].push_back(qSharedPointerCast<PointFeature>(feature));
if (std::find(fas.scales.begin(), fas.scales.end(), feature->scale) == fas.scales.end())
{
fas.scales.push_back(feature->scale);
}
}
}
break;
//Neighborhood features
case Feature::Type::NeighborhoodFeature:
{
//build the scaled feature list attached to the first cloud
if (feature->cloud1)
{
FeaturesAndScales& fas = cloudsWithScaledFeatures[feature->cloud1];
fas.neighborhoodFeaturesPerScale[feature->scale].push_back(qSharedPointerCast<NeighborhoodFeature>(feature));
++fas.featureCount;
if (std::find(fas.scales.begin(), fas.scales.end(), feature->scale) == fas.scales.end())
{
fas.scales.push_back(feature->scale);
}
}
//build the scaled feature list attached to the second cloud (if any)
if (feature->cloud2 && feature->cloud2 != feature->cloud1 && feature->op != Feature::NO_OPERATION)
{
FeaturesAndScales& fas = cloudsWithScaledFeatures[feature->cloud2];
fas.neighborhoodFeaturesPerScale[feature->scale].push_back(qSharedPointerCast<NeighborhoodFeature>(feature));
++fas.featureCount;
if (std::find(fas.scales.begin(), fas.scales.end(), feature->scale) == fas.scales.end())
{
fas.scales.push_back(feature->scale);
}
}
}
break;
//Context-based features
case Feature::Type::ContextBasedFeature:
{
//build the scaled feature list attached to the second cloud (the 'context' cloud)
if (feature->cloud2)
{
FeaturesAndScales& fas = cloudsWithScaledFeatures[feature->cloud2];
fas.contextBasedFeaturesPerScale[feature->scale].push_back(qSharedPointerCast<ContextBasedFeature>(feature));
++fas.featureCount;
if (std::find(fas.scales.begin(), fas.scales.end(), feature->scale) == fas.scales.end())
{
fas.scales.push_back(feature->scale);
}
}
}
break;
default:
assert(false);
break;
}
}
catch (const std::bad_alloc&)
{
errorStr = "Not enough memory";
return false;
}
}
}
bool success = true;
//if we have scaled features
if (!cloudsWithScaledFeatures.empty())
{
//for each cloud
for (QMap<ccPointCloud*, FeaturesAndScales>::iterator it = cloudsWithScaledFeatures.begin(); success && it != cloudsWithScaledFeatures.end(); ++it)
{
FeaturesAndScales& fas = it.value();
ccPointCloud* sourceCloud = it.key();
//sort the scales
std::sort(fas.scales.begin(), fas.scales.end());
//get the octree
ccOctree::Shared octree = sourceCloud->getOctree();
if (!octree)
{
ccLog::Print(QString("Computing octree of cloud %1 (%2 points)").arg(sourceCloud->getName()).arg(sourceCloud->size()));
octree = sourceCloud->computeOctree(progressCb);
if (!octree)
{
errorStr = "Failed to compute octree (not enough memory?)";
return false;
}
}
//now extract the neighborhoods from the biggest to the smallest scale
double largetScale = fas.scales.back();
PointCoordinateType largestRadius = static_cast<PointCoordinateType>(largetScale / 2); //scale is the diameter!
unsigned char octreeLevel = octree->findBestLevelForAGivenNeighbourhoodSizeExtraction(largestRadius);
unsigned pointCount = corePoints.size();
QString logMessage = QString("Computing %1 features on cloud %2\n(core points: %3)").arg(fas.featureCount).arg(sourceCloud->getName()).arg(pointCount);
if (progressCb)
{
progressCb->setMethodTitle("Compute features");
progressCb->setInfo(qPrintable(logMessage));
}
ccLog::Print(logMessage);
CCLib::NormalizedProgress nProgress(progressCb, pointCount);
QMutex mutex;
#ifndef _DEBUG
#if defined(_OPENMP)
#pragma omp parallel for
#endif
#endif
for (int i = 0; i < static_cast<int>(pointCount); ++i)
{
//spherical neighborhood extraction structure
CCLib::DgmOctree::NearestNeighboursSphericalSearchStruct nNSS;
{
nNSS.level = octreeLevel;
nNSS.queryPoint = *corePoints.cloud->getPoint(i);
nNSS.prepare(largestRadius, octree->getCellSize(nNSS.level));
octree->getTheCellPosWhichIncludesThePoint(&nNSS.queryPoint, nNSS.cellPos, nNSS.level);
octree->computeCellCenter(nNSS.cellPos, nNSS.level, nNSS.cellCenter);
}
//we extract the point's neighbors
unsigned kNN = octree->findNeighborsInASphereStartingFromCell(nNSS, largestRadius, true);
if (kNN != 0)
{
nNSS.pointsInNeighbourhood.resize(kNN);
//for each scale (from the largest to the smallest)
for (size_t scaleIndex = 0; scaleIndex < fas.scales.size(); ++scaleIndex)
{
if (scaleIndex != 0)
{
double radius = fas.scales[fas.scales.size() - 1 - scaleIndex] / 2; //scale is the diameter!
double sqRadius = radius * radius;
//remove the farthest points
for (; kNN > 0; --kNN)
{
if (nNSS.pointsInNeighbourhood[kNN - 1].squareDistd <= sqRadius)
{
break;
}
}
if (kNN == 0)
{
//no need to go further
break;
}
nNSS.pointsInNeighbourhood.resize(kNN);
}
//Point features
for (PointFeature::Shared& feature : fas.pointFeaturesPerScale[fas.scales[scaleIndex]])
{
if (feature->cloud1 == sourceCloud && feature->statSF1 && feature->field1)
{
double outputValue = 0;
if (!feature->computeStat(nNSS.pointsInNeighbourhood, feature->field1, outputValue))
{
//an error occurred
success = false;
break;
}
ScalarType v1 = static_cast<ScalarType>(outputValue);
feature->statSF1->setValue(i, v1);
}
if (feature->cloud2 == sourceCloud && feature->statSF2 && feature->field2)
{
assert(feature->op != Feature::NO_OPERATION);
double outputValue = 0;
if (!feature->computeStat(nNSS.pointsInNeighbourhood, feature->field2, outputValue))
{
//an error occurred
success = false;
break;
}
ScalarType v2 = static_cast<ScalarType>(outputValue);
feature->statSF2->setValue(i, v2);
}
}
//Neighborhhod features
for (NeighborhoodFeature::Shared& feature : fas.neighborhoodFeaturesPerScale[fas.scales[scaleIndex]])
{
if (feature->cloud1 == sourceCloud && feature->sf1)
{
double outputValue = 0;
if (!feature->computeValue(nNSS.pointsInNeighbourhood, nNSS.queryPoint, outputValue))
{
//an error occurred
success = false;
break;
}
ScalarType v1 = static_cast<ScalarType>(outputValue);
feature->sf1->setValue(i, v1);
}
if (feature->cloud2 == sourceCloud && feature->sf2)
{
assert(feature->op != Feature::NO_OPERATION);
double outputValue = 0;
if (!feature->computeValue(nNSS.pointsInNeighbourhood, nNSS.queryPoint, outputValue))
{
//an error occurred
success = false;
break;
}
ScalarType v2 = static_cast<ScalarType>(outputValue);
feature->sf2->setValue(i, v2);
}
}
//Context-based features
for (ContextBasedFeature::Shared& feature : fas.contextBasedFeaturesPerScale[fas.scales[scaleIndex]])
{
if (feature->cloud2 == sourceCloud && feature->sf)
{
ScalarType outputValue = 0;
if (!feature->computeValue(nNSS.pointsInNeighbourhood, nNSS.queryPoint, outputValue))
{
//an error occurred
success = false;
break;
}
feature->sf->setValue(i, outputValue);
}
}
if (!success)
{
break;
}
}
} //for each scale
if (progressCb)
{
mutex.lock();
bool cancelled = !nProgress.oneStep();
mutex.unlock();
if (cancelled)
{
//process cancelled by the user
ccLog::Warning("Process cancelled");
errorStr = "Process cancelled";
success = false;
break;
}
}
} //for each point
} //for each cloud
}
for (const Feature::Shared& feature : features)
{
//we have to 'finish' the process for scaled features
if (feature->scaled() && !feature->finish(corePoints, errorStr))
{
return false;
}
}
return success;
}
bool Tools::RandomSubset(ccPointCloud* cloud, float ratio, CCLib::ReferenceCloud* inRatioSubset, CCLib::ReferenceCloud* outRatioSubset)
{
if (!cloud)
{
ccLog::Warning("Invalid input cloud");
return false;
}
if (!inRatioSubset || !outRatioSubset)
{
ccLog::Warning("Invalid input refence clouds");
return false;
}
if (inRatioSubset->getAssociatedCloud() != cloud || outRatioSubset->getAssociatedCloud() != cloud)
{
ccLog::Warning("Invalid input reference clouds (associated cloud is wrong)");
return false;
}
if (ratio < 0.0f || ratio > 1.0f)
{
ccLog::Warning(QString("Invalid parameter (ratio: %1)").arg(ratio));
return false;
}
unsigned inSampleCount = static_cast<unsigned>(floor(cloud->size() * ratio));
assert(inSampleCount <= cloud->size());
unsigned outSampleCount = cloud->size() - inSampleCount;
//we draw the smallest population (faster)
unsigned targetCount = inSampleCount;
bool defaultState = false;
if (outSampleCount < inSampleCount)
{
targetCount = outSampleCount;
defaultState = true;
}
//reserve memory
std::vector<bool> pointInsideRatio;
try
{
pointInsideRatio.resize(cloud->size(), defaultState);
}
catch (const std::bad_alloc&)
{
ccLog::Warning("Not enough memory");
return false;
}
if (!inRatioSubset->reserve(inSampleCount) || !outRatioSubset->reserve(outSampleCount))
{
ccLog::Warning("Not enough memory");
inRatioSubset->clear();
outRatioSubset->clear();
return false;
}
//randomly choose the 'in' or 'out' indexes
int randIndex = 0;
unsigned randomCount = 0;
while (randomCount < targetCount)
{
randIndex = ((randIndex + std::rand()) % cloud->size());
if (pointInsideRatio[randIndex] == defaultState)
{
pointInsideRatio[randIndex] = !defaultState;
++randomCount;
}
}
//now dispatch the points
{
for (unsigned i = 0; i < cloud->size(); ++i)
{
if (pointInsideRatio[i])
inRatioSubset->addPointIndex(i);
else
outRatioSubset->addPointIndex(i);
}
assert(inRatioSubset->size() == inSampleCount);
assert(outRatioSubset->size() == outSampleCount);
}
return true;
}
Reference in New Issue View Git Blame Copy Permalink