opencv - BOW with more 2-classes SVMs of one multiclass SVM -
because opencv's forum has problem, want post question here too.
in example of bow opencv, there trained 2-classes svm classifier each class. why isn't use multi-class svm instead? if this, then, supposing have x classes, have load x svms , predict x times image.
because code large , uses large dataset, takes long doing research.
[ have managed start training first svm classifier opencv's bow example, says:
143 positive training samples; 2356 negative training samples
i wondering if not bad classifier... maybe train_auto
find needed parameters not bad. think? ]
Comments
Post a Comment