Sorry noob asked this question, but because I really need some guidance on using Multi probe LSH quite urgently, so I did not do a lot of research myself. I understand that there is access to LSLKIT for lib that implemented this algorithm, but itโs hard for me to figure out how to use it. Right now I have several thousand vector values โโof 296, each of which represents an image. A vector is used to request a custom input image to get the most similar image. The method that I used to determine the distance between the vector is the Euclidean distance.
I know this may be a pretty empty question, but did you know that I have knowledge on how I should use the multi-probe LSH? I am very grateful for any answer or answer.
- update -
Tried to create a model for my data with fitdata tools provided, however this does not seem to apply to my file. The format that I used for input is presented in this format, the size of the float: 4, the amount of data: 20, the dimension: 297 and my array with a floating matrix 297. However, it gives me this error
gsl: init_source.c:29: ERROR: matrix dimension n1 must be positive integer Default GSL error handler invoked. Aborted
Do you have any idea how to create an input for fitdata?
- update -
Sorry for the last update by trying lsh. You can use text2bin to format data for fitdata. The text file contains a function vector image or audio file with each line representing the vector. After that, use mplsh-tun to get the M and W. parameters. To build the index, you can use the scan tool to try the set of required queries, and you can use mplsh-run to get the index. Now I'm trying to figure out how to use the index and how to link the library with my encoding. Does any body have an idea about this?
c ++ algorithm image-processing machine-learning hash
Yijinsei
source share