1
$\begingroup$

I'm doing image classification by extracting SIFT features, clustering them and then finding BOVW histogram and classifying.

I have around 180 training images from which I'm extracting SIFT descriptors. I need to cluster these features using k-Means clustering. Now, among the SIFT descriptors, some are duplicate. Before applying K-Means, should I remove these duplicate vectors or should I not? Or does it not make any difference?

Thanks!

$\endgroup$
1
$\begingroup$

It does make a difference.

If you have many duplicates, you can merge them into weighted vectors. It's straightforward to add to existing code.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.