So after training and saving my model. I tried to load it again only to find that there's no load function in the java wrappers. After some googling I found that this is a bug. It looks like the python wrappers are fixed but no word on the java ones.
So is there a work around for this? Because training the model each run is a bit too much. I'm using OpenCV 3.1.0
the python fix also applies to java.
(maybe you need to update your opencv libs)
if you compile the latest master branch, you should have:
//
// C++: static Ptr_SVM load(String filepath)
//
//javadoc: SVM::load(filepath)
public static SVM load(String filepath)
{
SVM retVal = new SVM(load_0(filepath));
return retVal;
}
in build/modules/java/ml+SVM.java, and you would use it like:
SVM svm = SVM.load("my.yml");
That's what I also though but I downloaded the latest release from opencv.org (3.1.0) and used the prebuilt jar file. I just redownloaded for a sanity check but still no load function.
The load function is a function from the algorithmclass. The Algorithm class in java doesn't contain a load function.
unfortunately, the prebuilt libs always lag behind master.
Algorithm::load() is a template function in c++, that's why you need this overload to use it from python or java
any chance, that you can rebuild the libs locally ? (needs cmake, ant, jdk and a c++ compiler)
Hmm yeah that's something I'm going to do another day then. Never done that before but I found this tutorial. So it shouldn't be to hard I guess.
omg I finally got it compiled. And now I get errors on smv.save.
Exception in thread "main" java.lang.Exception: unknown exception Yeah that's a clear error message /s
Asked: 2016-04-10 07:49:30 -0600
Seen: 847 times
Last updated: Apr 11 '16
