I am currently working on an image classification project ,where i need to use bag of features model with libsvm(HIK kernel) for classification.I have extracted sift feature and created a visual vocabulary,my fetaures are double values.But LIBSVM works with only integer values.Can anyone please help me with this?
libsvm doesn't work with only integer values -- in fact, you're supposed to scale your features to lie within [-1,1] or [0,1] before applying SVM's typically. Read their guide. They also have some tricks for dealing with categorical features as well.
If you want to get finitely many values which the feature can take, use quantization.