Revision history [back]

yes, createLBPHFaceRecognizer() is missing in the java api, it's a problem with the current generated wrappers (it can't handle cv::Ptr<anything> properly and just skips it).

so, what are the options ?

1. wait a bit. there's a pull request in the pipeline, that should fix it.

2. roll your own with a bit of jni (see here and here)

3. fall back to javacv (bah)

4. no idea , what you need it for, but as a proof of concept, a straight norm(a,b,L2) does surprisingly well (even beats eigenfaces unless you do heavy preprocessing)

yes, createLBPHFaceRecognizer() is missing in the java api, it's a problem with the current generated wrappers (it can't handle cv::Ptr<anything> properly and just skips it).

so, what are the options ?

1. wait a bit. there's a pull request in the pipeline, that should fix it.

2. roll your own with a bit of some jni (see here and here)

3. fall back to javacv (bah)

4. no idea , what you need it for, but as a proof of concept, a straight norm(a,b,L2) does surprisingly well (even beats eigenfaces unless you do heavy preprocessing)

yes, createLBPHFaceRecognizer() (as well as the other 2 create** methods) is missing in the java api, it's a problem with the current generated wrappers (it can't handle cv::Ptr<anything> properly and just skips it).

so, what are the options ?

1. wait a bit. there's a pull request in the pipeline, that should fix it.

2. roll your own with some jni (see here and here)

3. fall back to javacv (bah)

4. no idea , what you need it for, but as a proof of concept, a straight norm(a,b,L2) does surprisingly well (even beats eigenfaces unless you do heavy preprocessing)

yes, createLBPHFaceRecognizer() (as well as the other 2 create** methods) is missing in the java api, it's a problem with the current generated wrappers (it can't handle cv::Ptr<anything> properly and just skips it).

so, what are the options ?

1. wait a bit. there's a pull request in the pipeline, that should fix it.

2. roll your own with some jni (see here and here)

3. fall back to javacv (bah)

4. no idea , what you need it for, but as a proof of concept, a straight norm(a,b,L2) does surprisingly well (even beats eigenfaces unless you do heavy preprocessing)

5. outsource it to a server. usually, doing the recognition on your phone is a bad idea, as you probably want a huge database, and a consistent one for all users of your app.