c++ - Is there any checklist to avoid crashes when using OpenCV SVM? -
i' stepping through svm in opencv 2.4 slowly. 1 crash there list of suggestions, how check data befode feeding train , predict functions?
currently have crashes when calling predict , can't figure out why.
this section of code should relevant:
row = cv::mat::zeros(1, 256, cv_32fc1); ( std::map<int, int>::iterator fit = tmp.begin(); fit != tmp.end(); fit ++ ) { row.at<float>(0, fit->first) = fit->second; } float result = svm.predict(row);
seems not documented stumbling block. when create cvsvm - instance on heap, doesn't crash more.
i used create global instance ov cvsvm in test code, , assign actual data later:
cvsvm _svm; //... code , scopes //training _svm = cvsvm (trainingdata, trainingclasses, cv::mat(), cv::mat(), param); //... code , scopes //testing float l = _svm.predict(testdata)
it crashed in predict, because labels mat vas not initialized
Comments
Post a Comment