How to use VLfeat to do VLAD-k encoding?

Hi everyone, I want to use VLfeat library to do VLAD-k encoding of image features. They have a simple tutorial online at here (<http://www.vlfeat.org/overview/encodings.html)>. But the tutorial is a little vague for me to follow. In addition, the example is for VLAD with nearest neighbor, not VLAD-k. So could anyone write an example of VLAD-k using VLfeat for me? Thanks a lot.

[Yi Zhu]

VLAD-k exmaple I wrote, is this correct? Especially for the assignment step? Thanks.

numFeatures = 5000 ;
dimension = 2 ;
data = rand(dimension,numFeatures) ;      % Data is used to train the VLAD encoding

numDataToBeEncoded = 1000;
dataToBeEncoded = rand(dimension,numDataToBeEncoded);

numClusters = 30 ;
centers = vl_kmeans(data, numClusters);
kdtree = vl_kdtreebuild(centers) ;
nn = vl_kdtreequery(kdtree, centers, dataToBeEncoded,'NUMNEIGHBORS', k) ;    % k is the number of nearest neighbors

assignments = zeros(numClusters,numDataToBeEncoded);
assignments(sub2ind(size(assignments), nn, 1:length(nn))) = 1;      % is here still correct for VLAD-k? nn is a matrix now

enc = vl_vlad(dataToBeEncoded,centers,assignments);

Answer 1

[ahmed shahin]

Dear Zhu

what is the value of k?