Ask Your Question
0

which types of descriptors opencv use for machine learning( local or global)

asked 2016-11-28 15:39:46 -0600

jack1 gravatar image

updated 2016-11-28 16:22:44 -0600

I do some researches about descriptors used in classification algorithms. I use opencv such a library so i should know how opencv machine learning algorithms work and which descriptors they use by default( local like sift and surf or global like color and shape) and if i can use other types of descriptors. Thanks in advance

edit retag flag offensive close merge delete

Comments

i also think, you're somehow asking the wrong question.

what's your context ? what problem are you trying to solve ?

berak gravatar imageberak ( 2016-11-29 01:09:47 -0600 )edit

I would like to use SIFT/ SURT/HOG as features for machine learning algorithm like SVM/ANN

jack1 gravatar imagejack1 ( 2016-11-29 05:22:46 -0600 )edit

sure, what is stopping you ?

(for sift/surf, you will probably have to use some BOW approach (because descriptor count varies), while for HOG, you'll need a way to get tons of equally cropped positives/negatives)

berak gravatar imageberak ( 2016-11-29 05:38:43 -0600 )edit

I use Sift/SURF for the first time and don't know a lot about them. please can you give me an example in java that use theses descriptors for classification

jack1 gravatar imagejack1 ( 2016-11-29 05:48:09 -0600 )edit

sorry, cannot help much with java, atm. (not even sure, if the BOW methods are available from java)

berak gravatar imageberak ( 2016-11-29 05:55:12 -0600 )edit

Ok thank you a lot

jack1 gravatar imagejack1 ( 2016-11-29 07:14:57 -0600 )edit

1 answer

Sort by ยป oldest newest most voted
1

answered 2016-11-28 17:30:00 -0600

Tetragramm gravatar image

I think your question is wrong. The tools in the Machine Learning module of OpenCV don't use anything in particular. Or rather, they use what you give them.

The only learning technique that uses descriptors is the BoW stuff, and that just uses whichever descriptor type you give it.

edit flag offensive delete link more

Comments

Thank you, so how to use SIFT/SURF or any other descriptors as features for machine learning

jack1 gravatar imagejack1 ( 2016-11-29 05:14:35 -0600 )edit

I dunno. It depends on what you're doing with them. You probably use them as input data rather than output data. So for an SVM, they would be the samples variable of TrainData::create. It would probably help to read the tutorial HERE. It's guaranteed to be more accurate than I am.

Tetragramm gravatar imageTetragramm ( 2016-11-29 10:39:34 -0600 )edit

Question Tools

1 follower

Stats

Asked: 2016-11-28 15:39:46 -0600

Seen: 268 times

Last updated: Nov 28 '16