NormalBayesClassifier classifier is an abstract class

asked 2014-09-03 16:37:05 -0600

kenilworth
21 ●1 ●5

updated 2014-09-04 02:14:00 -0600

berak
32993 ●7 ●81 ●312

I'm trying to figure out how to use the Normal Bayes Classifier as explained here.

Mat trainingData;
Mat trainingLabels;
Mat evalData;
Mat results;

cv::ml::NormalBayesClassifier classifier;
//Train classifier...
classifier.train(trainingData, trainingLabels);

//Evaluate classifier...
classifier.predict(evalData,&results);

But I get the following build error: cv::ml::NormalBayesClassifier classifier is an abstract class

I tried making a subclass of NormalBayesClassifier and using that instead but I didn't manage.. What am I doing wrong?

edit retag flag offensive close merge delete

add a comment

1 answer

Sort by » oldest newest most voted

answered 2014-09-04 01:51:38 -0600

berak
32993 ●7 ●81 ●312

updated 2014-09-04 02:15:38 -0600

since you're obviously running opencv3.0, you can no more create it on the 'stack' (like it was in 2.4.9). it does not have a public constructor (that's your 'abstract class' error above)

instead, you will have to use a factory function, returning a cv::Ptr :

cv::Ptr<cv::ml::NormalBayesClassifier>  classifier = cv::ml::NormalBayesClassifier::create();

//Train classifier...
classifier->train(trainingData, cv::ml::ROW_SAMPLE, trainingLabels);
// note: it needs a flag now, if a feature is on a row or a column.

//Evaluate classifier...
classifier->predict(evalData,results); 
// note: i skipped the address-operator on results,
// which probably was wrong in the 1st place

also, better look at the 3.0 documentation, i.e. the ml module got some huge overhaul lately.