QList<Model::Node*> LoopStatement::findSymbols(const QRegExp& symbolExp,Model::Node* source, FindSymbolMode mode,
bool exhaustAllScopes)
{
QList<Model::Node*> symbols;
if (condition()) symbols << condition()->findSymbols(symbolExp, source, SEARCH_DOWN, false);
if (initStep()) symbols << initStep()->findSymbols(symbolExp, source, SEARCH_DOWN, false);
if (updateStep()) symbols << updateStep()->findSymbols(symbolExp, source, SEARCH_DOWN, false);
if (exhaustAllScopes || symbols.isEmpty())
symbols << Node::findSymbols(symbolExp, source, mode, exhaustAllScopes);
return symbols;
}
void eapAddStep(char *pattern)
/* eapAddStep - Add a step to eapStep and related tables. This is just a small shortcut for doing
* it in SQL. You can only add steps defined in C code.. */
{
struct sqlConnection *conn = eapConnectReadWrite();
int i;
for (i=0; i<ArraySize(steps); ++i)
{
struct stepInit *init = &steps[i];
if (wildMatch(pattern, init->name))
{
if (clList)
puts(init->name);
else
initStep(conn, init);
}
}
}
void ParticleEmitter::step(ParticleData& particles, PxReal dt, const PxVec3& externalAcceleration, PxReal maxParticleVelocity)
{
initStep(particles, dt);
stepInternal(particles, dt, externalAcceleration, maxParticleVelocity);
finalizeStep();
}
int main(int argc, char ** argv)
{
const std::string patchParamsFilename = argv[1];
const std::string testDataFilename = argv[2];
const std::string valveDataFilename = "/home/om0000/ValveTracking/TrainingData/MV-Planes.hdf5";
PatchParams params(patchParamsFilename);
TestData::Pointer testData = TestData::Initialise(testDataFilename);
LengthData::Pointer lengthData = LengthData::Load("/home/om0000/ValveTracking/TrainingData/Lengths");
lengthData->Prepare(testData->GetId());
std::cout << "Loading Plane Data" << std::endl;
ValveTrainingData::Pointer valveData = ValveTrainingData::Load(valveDataFilename);
std::cout << "Loading PatchData" << std::endl;
//PatchTrainingData::Pointer patchData = PatchTrainingData::Load(patchParamsFilename);
PatchTrainingData::Pointer patchData = PatchTrainingData::New();
VectorType startPoint, startNormal;
for(unsigned int runs = 0; runs < 1; runs++)
{
for(params.timeStep = 0; params.timeStep < params.numberOfTimeSteps; params.timeStep++)
{
std::cout << "Working on: " << params.timeStep << std::endl;
// load the input points and create the bounding box
MatrixType pointsData;
valveData->GetTrainingPoints(testData->GetId(), params.timeStep, pointsData);
std::vector<MatrixType> pointsData2;
valveData->GetTrainingPoints(testData->GetId(), params.timeStep, pointsData2);
vtkBoundingBox boundingBox;
computeBoundingBox(pointsData, boundingBox);
std::cout << "Loading / Training Classifiers" << std::endl;
ClassifierMap classifiers;
loadOrTrainClassifiers(params, 1000, patchData, classifiers);
testData->Prepare(params.timeStep, boundingBox);
testData->SaveImageGroup("2C", "MV-2C");
// compute the distribution for the plane and get the mean for initialisation
MatrixType planeData;
valveData->GetTrainingData(testData->GetId(), params.timeStep, planeData);
PlaneCovarianceFilterType::Pointer planeDistribution = PlaneCovarianceFilterType::New();
computePlaneCovariance(planeData, planeDistribution);
TestData::VectorType point, normal;
if(params.timeStep == 0 && runs == 0)
{
for(unsigned int i = 0; i < 3; i++)
{
point(i) = planeDistribution->GetMean()[i];
normal(i) = planeDistribution->GetMean()[i+3];
}
//startingPlane(pointsData2, point, normal);
}
else
{
point = startPoint;
normal = startNormal;
}
OptData * optData = new OptData;
optData->testData = testData;
optData->classifiers = classifiers;
optData->params = params;
optData->lengths = lengthData;
optData->planeMember = planeDistribution;
nlopt::opt opt(nlopt::LN_BOBYQA, 5);
opt.set_min_objective(f, (void*) optData);
opt.set_xtol_rel(1e-4);
opt.set_maxeval(100);
std::vector<double> initStep(5);
double pointStep = atof(argv[3]);
double normalStep = atof(argv[4]);
for(unsigned int i = 0; i < 3; i++)
{
initStep[i] = pointStep;
}
initStep[3] = normalStep;
initStep[4] = normalStep;
opt.set_default_initial_step(initStep);
std::vector<double> xStart;
convert_plane(point, normal, xStart);
double minf;
nlopt::result result = opt.optimize(xStart,minf);
std::cout << "Final: " << result << std::endl;
MatrixType initialPlane(1,6), finalPlane(1,6);
VectorType finalPoint, finalNormal;
convert_x(xStart, finalPoint, finalNormal);
imageCost(params, testData, finalPoint, finalNormal, classifiers, lengthData, true);
for(unsigned int i = 0; i < 3; i++)
{
finalPlane(0,i) = finalPoint(i);
finalPlane(0,i+3) = finalNormal(i);
}
std::cout << finalPlane << std::endl;
for(unsigned int i = 0; i < 6; i++)
{
initialPlane(0,i) = planeDistribution->GetMean()[i];
}
startPoint = finalPoint;
startNormal = finalNormal;
MatrixType gt;
valveData->GetTestData(testData->GetId(), params.timeStep, gt);
VectorType gtPoint, gtNormal;
for(unsigned int i = 0; i < 3; i++)
{
gtPoint(i) = gt(0,i);
gtNormal(i) = gt(0,i+3);
}
std::vector<ImageType::Pointer> vimages;
testData->GetImages(vimages);
ResultViewer::Pointer viewer = ResultViewer::New();
viewer->SetImages(vimages);
viewer->SetBoundingBox(boundingBox);
viewer->SetPlane(finalPoint, finalNormal);
viewer->SetStartPlane(gtPoint, gtNormal);
viewer->SetUp();
std::stringstream ss;
ss << params.name << "/" << testData->GetId() << "-" << params.timeStep << "-image.png";
viewer->Save(ss.str());
//viewer->View();
//testData->SaveImageGroup("2C", "2C");
//testData->SaveImageGroup("3C", "3C");
//testData->SaveImageGroup("4C", "4C");
//savePlane(initialPlane, "initialPlane.vtk");
//savePlane(finalPlane, "finalPlane.vtk");
}
}
return 0;
}
