Opencv does have support SVM accuracy?
I had using opencv 3.3 to developed App Motobike plate recognition on C++. I'd using over 600 image to train. Then i used 300 other image for testing training data. How does i can find the result accuracy of testing from svm? Or how i can find any character detect not correct without manual?
what are you doing there, exactly ? are you trying to read the letters (or only - plate or not) ? do you have any "ground truth" data for your test set ?
You need manually labelling each image (ground truth or not + exact characters in ground truth) in order to compare SVM plate recognition testing results combined with OCR recognition to your own manual labels
@berak: im trying read letters on plate. im using plate image from phone camera
@carton99: your mean that SVM cant show accuracy without label. Each image to checking dont have percent accuracy if we don't set label before?
@vbc: if you want evaluate your method you need to have label for your testing group image.
Nevertheless, I think you mix with the heuristic given in output by the SVM which give a value on a probability to belonging to your objet (plate).