[doc] Fisher/gmm: data is expected to be row major about vlfeat

[vedaldi]

Dear Luk,

for matrices in VLFeat we use column-major storage (mostly to be compatible with MATLAB and other numerical packages) and row-major for images (for compatibility with image processing and graphical libraries). In this sense, the documentation is correct in saying that features are stored one per column, although admittedly in C very often one thinks "row major" a priori. I will try to think of a way to clarify this better in the docs.

Thank you for your note!

Andrea

On 31 Aug 2013, at 12:50, luk [email protected] wrote:

Hi,
The documentation indicates the data matrix has one feature vector per column

[...] data has dimension rows and numData columns.

However, afaic througout fisher.c and gmm.c data access seems to be row major and thus assumes a feature vector per row for example

for(dim = 0; dim < dimension; dim++) {
TYPE diff = data[i_d_dimension + dim] - means[i_cl_dimension + dim] ;
I would propose to adapt the documentation accordingly :)

—
Reply to this email directly or view it on GitHub.

[luk]

Thanks for your rapid answer! It's true row/col major and the concatentation of vector data are two different things. Maybe adding your explanetory words from this issue to the "General support functionalities" section of the doc would alredy help someone not to make the same mistake as I did :)

Thank you very much for your work!
Lukas