rmsalinas / fbow

FBOW (Fast Bag of Words) is an extremmely optimized version of the DBow2/DBow3 libraries.
559 stars 141 forks source link

Different vocabularies generated from same data #24

Closed dukeNashor closed 4 years ago

dukeNashor commented 5 years ago

I tested fbow on same data for several times. It is worth noticing that different instances operating on the same data generated different vocabularies, while dbow3 generated the exact same vocabulary. is this normal?

dukeNashor commented 5 years ago

After some debugging, i noticed that the feature points in initial cluster may (almost always, in my case) duplicate; after changing the initial cluster selection part, everything worked fine.

RashidLadj commented 4 years ago

Hi @dukeNashor

I can't quite understand the problem, I took a look at your repository but you haven't updated it on git, can you tell me where exactly is the problem?

Thank you.

dukeNashor commented 4 years ago

Hi @dukeNashor

I can't quite understand the problem, I took a look at your repository but you haven't updated it on git, can you tell me where exactly is the problem?

Thank you.

nvm, the fbow's implementation for choosing the initial cluster centers is different from dbow's, which leads to the fact that different runs of fbow generate different vocabularies.