How to create a Superpoint vocabulary?

[antithing]

Hi, i am building a relocalization system using Superpoint deep features.

Can i just follow the steps in this repo, but use the 256 bit descriptor? Will this cause any issues? or will it 'just work?'

Thank you!

[antithing]

adding to this.. I have created a vocabulary successfully, but when load it and try to use it, I see that that :

_voc.getDescSize() == 1024

when i would expect 256 (as that is the cols of a superpoint descriptor).

I used the example code in the fbow repo, what have i done wrong?

Thanks!

[ThomasZiegler]

I came across the same problem with the descriptor size being four times bigger than expected. Reason can be seen here

fbow/src/vocabulary_creator.cpp

Line 26 in e148dbc

_descNBytes=features[0].cols* features[0].elemSize();

The return descriptor size is the number of descriptors times the number of bytes per descriptor. I.e. if the descriptor uses float (32 bits = 4 bytes).

[llllzq]

Do you meet the problem that the output of two pic fbow:fbow:score is 0?

[antithing]

Hi @ThomasZiegler @llllzq did you successfully train and run a superpoint vocabulary? If so would you be able to share your workflow? Thank you