generic-library/opencv
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

fixed a few VS2010 warnings in EM

Browse Source
This commit is contained in:
Vadim Pisarevsky 2012-04-06 10:04:22 +00:00
parent 2291faedd8
commit 0f1f7b0b4a
2 changed files with 9 additions and 9 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

16
modules/ml/src/em.cpp
View File

@ -44,7 +44,7 @@
namespace cv namespace cv
{ {
const float minEigenValue = 1.e-3; const float minEigenValue = 1.e-3f;
EM::Params::Params( int nclusters, int covMatType, int startStep, const cv::TermCriteria& termCrit, EM::Params::Params( int nclusters, int covMatType, int startStep, const cv::TermCriteria& termCrit,
const cv::Mat* probs, const cv::Mat* weights, const cv::Mat* probs, const cv::Mat* weights,
@ -124,7 +124,7 @@ int EM::predict(const cv::Mat& sample, cv::Mat* _probs, double* _likelihood) con
CV_Assert(sample.type() == CV_32FC1); CV_Assert(sample.type() == CV_32FC1);
int label; int label;
float likelihood; float likelihood = 0.f;
computeProbabilities(sample, label, _probs, _likelihood ? &likelihood : 0); computeProbabilities(sample, label, _probs, _likelihood ? &likelihood : 0);
if(_likelihood) if(_likelihood)
*_likelihood = static_cast<double>(likelihood); *_likelihood = static_cast<double>(likelihood);
@ -254,7 +254,7 @@ void preprocessProbability(cv::Mat& probs)
{ {
cv::max(probs, 0., probs); cv::max(probs, 0., probs);
const float uniformProbability = 1./probs.cols; const float uniformProbability = (float)(1./probs.cols);
for(int y = 0; y < probs.rows; y++) for(int y = 0; y < probs.rows; y++)
{ {
cv::Mat sampleProbs = probs.row(y); cv::Mat sampleProbs = probs.row(y);
@ -395,7 +395,7 @@ void EM::computeLogWeightDivDet()
for(int di = 0; di < covsEigenValues[clusterIndex].cols; di++) for(int di = 0; di < covsEigenValues[clusterIndex].cols; di++)
logDetCov += std::log(covsEigenValues[clusterIndex].at<float>(covMatType != EM::COV_MAT_SPHERICAL ? di : 0)); logDetCov += std::log(covsEigenValues[clusterIndex].at<float>(covMatType != EM::COV_MAT_SPHERICAL ? di : 0));
logWeightDivDet.at<float>(clusterIndex) = logWeights.at<float>(clusterIndex) - 0.5 * logDetCov; logWeightDivDet.at<float>(clusterIndex) = logWeights.at<float>(clusterIndex) - 0.5f * logDetCov;
} }
} }
@ -421,7 +421,7 @@ bool EM::doTrain(const cv::TermCriteria& termCrit)
if(startStep == EM::START_M_STEP) if(startStep == EM::START_M_STEP)
mStep(); mStep();
double trainLikelihood, prevTrainLikelihood; double trainLikelihood, prevTrainLikelihood = 0.;
for(int iter = 0; ; iter++) for(int iter = 0; ; iter++)
{ {
eStep(); eStep();
@ -489,7 +489,7 @@ void EM::computeProbabilities(const cv::Mat& sample, int& label, cv::Mat* probs,
Lval += w * val * val; Lval += w * val * val;
} }
CV_DbgAssert(!logWeightDivDet.empty()); CV_DbgAssert(!logWeightDivDet.empty());
Lval = logWeightDivDet.at<float>(clusterIndex) - 0.5 * Lval; Lval = logWeightDivDet.at<float>(clusterIndex) - 0.5f * Lval;
L.at<float>(clusterIndex) = Lval; L.at<float>(clusterIndex) = Lval;
if(Lval > L.at<float>(label)) if(Lval > L.at<float>(label))
@ -508,7 +508,7 @@ void EM::computeProbabilities(const cv::Mat& sample, int& label, cv::Mat* probs,
if(clusterIndex != label) if(clusterIndex != label)
partExpSum += expL.at<float>(clusterIndex); partExpSum += expL.at<float>(clusterIndex);
} }
factor = 1./(1 + partExpSum); factor = 1.f/(1 + partExpSum);
cv::exp(L - L.at<float>(label), expL); cv::exp(L - L.at<float>(label), expL);
@ -522,7 +522,7 @@ void EM::computeProbabilities(const cv::Mat& sample, int& label, cv::Mat* probs,
if(likelihood) if(likelihood)
{ {
// note likelihood = log (sum_j exp(L_ij)) - 0.5 * dims * ln2Pi // note likelihood = log (sum_j exp(L_ij)) - 0.5 * dims * ln2Pi
*likelihood = std::log(partExpSum + expL.at<float>(label)) - 0.5 * dim * CV_LOG2PI; *likelihood = std::log(partExpSum + expL.at<float>(label)) - (float)(0.5 * dim * CV_LOG2PI);
} }
} }

2
modules/ml/test/test_emknearestkmeans.cpp
View File

@ -490,7 +490,7 @@ protected:
Mat firstResult(samples.rows, 1, CV_32FC1); Mat firstResult(samples.rows, 1, CV_32FC1);
for( int i = 0; i < samples.rows; i++) for( int i = 0; i < samples.rows; i++)
firstResult.at<float>(i) = em.predict( samples.row(i) ); firstResult.at<float>(i) = (float)em.predict( samples.row(i) );
// Write out // Write out
string filename = tempfile() + ".xml"; string filename = tempfile() + ".xml";